MALTE MEINSHAUSEN , BILL HARE , TOM M.L. WIGLEY , DETLEF … · 2018. 7. 7. · MULTI-GAS EMISSIONS...

MULTI-GAS EMISSIONS PATHWAYS TO MEET CLIMATE TARGETS

MALTE MEINSHAUSEN1,3, BILL HARE2, TOM M.L. WIGLEY3,DETLEF VAN VUUREN4, MICHEL G.J. DEN ELZEN4 and ROB SWART5

1Swiss Federal Institute of Technology (ETH Zurich), Environmental Physics,Environmental Science Department, Ramistrasse 101, CH-8092 Zurich, Switzerland

Email: [email protected] Scientist, Potsdam Institute for Climate Impact Research (PIK), Telegrafenberg A31,

D-14412 Potsdam, Germany3National Center for Atmospheric Research, NCAR, P.O. Box 3000, Boulder, CO 80307,

Colorado, United States4Netherlands Environmental Assessment Agency (MNP), 3720 BA Bilthoven,

The Netherlands5EEA European Topic Center for Air and Climate Change (ETC/ACC), MNP

3720 BA Bilthoven, The Netherlands

Abstract. So far, climate change mitigation pathways focus mostly on CO2 and a limited number ofclimate targets. Comprehensive studies of emission implications have been hindered by the absence ofa flexible method to generate multi-gas emissions pathways, user-definable in shape and the climatetarget. The presented method ‘Equal Quantile Walk’ (EQW) is intended to fill this gap, buildingupon and complementing existing multi-gas emission scenarios. The EQW method generates newmitigation pathways by ‘walking along equal quantile paths’ of the emission distributions derivedfrom existing multi-gas IPCC baseline and stabilization scenarios. Considered emissions includethose of CO2 and all other major radiative forcing agents (greenhouse gases, ozone precursors andsulphur aerosols). Sample EQW pathways are derived for stabilization at 350 ppm to 750 ppm CO2

concentrations and compared to WRE profiles. Furthermore, the ability of the method to analyzeemission implications in a probabilistic multi-gas framework is demonstrated. The probability ofovershooting a 2 ◦C climate target is derived by using different sets of EQW radiative forcing peakingpathways. If the probability shall not be increased above 30%, it seems necessary to peak CO2

equivalence concentrations around 475 ppm and return to lower levels after peaking (below 400 ppm).EQW emissions pathways can be applied in studies relating to Article 2 of the UNFCCC, for theanalysis of climate impacts, adaptation and emission control implications associated with certainclimate targets. See www.simcap.org for EQW-software and data.

1. Introduction

Ten years after its entry into force, the United Nations Framework Convention onClimate Change (UNFCCC) has been ratified by 188 countries.1 It calls for the pre-vention of ‘dangerous anthropogenic interference with the climate system’ (Article2). In order to study the transient climate impacts of human-induced greenhousegas (GHG) emissions and its implications for emission control policies, multi-gasemissions pathways that capture a wide range of intervention and non-interventionemission futures are required.

Climatic Change (2006) 75: 151–194DOI: 10.1007/s10584-005-9013-2 c© Springer 2006

brought to you by COREView metadata, citation and similar papers at core.ac.uk

provided by RERO DOC Digital Library

https://core.ac.uk/display/159154454?utm_source=pdf&utm_medium=banner&utm_campaign=pdf-decoration-v1

152 M. MEINSHAUSEN ET AL.

The aim of this study is to present a method that can simultaneously meet threegoals relevant to studies relating to Article 2.

• The first goal is to generate multi-gas emissions pathways consistent withthe range of climate policy target indicators under discussion. The targetparameter and its level can be freely selected. Examples of target parametersinclude CO2 concentrations, radiative forcing, global mean temperatures orsea level rise.

• The second goal is that the multi-gas pathways generated should have a treat-ment of non-CO2 gases and radiative forcing agents that is consistent withthe range of multi-gas scenarios in the literature. The inclusion of a non-CO2

component in the newly created emissions pathways might significantly im-prove on mitigation pathways generated in the past but without the necessityof a comprehensive analysis of mitigation options across energy, agriculture,and other sectors. Several studies have shown that it is important to take intoaccount the full range of greenhouse gases including, but not limited to, thesix greenhouse gases and gas groups controlled by the Kyoto Protocol both foreconomic cost-effectiveness and climatic reasons (Reilly et al., 1999; Hansenet al., 2000; Manne and Richels, 2001; Sygna et al., 2002; Eickhout et al.,2003; van Vuuren et al., 2003). However, until recently, most studies havefocused on CO2 only.

• The third goal is to create a method to generate multi-gas pathways for user-specified climate targets. Developing a flexible method, rather than only a lim-ited number of mitigation pathways, has significant advantages. For example,it can facilitate a comprehensive exploration of the emission implications ofcertain climate targets, given our scientific uncertainties in the main climatesystems components, such as climate sensitivity and ocean diffusivity.

There are two broad classifications of emissions pathways: a non-interventionist(baseline) path or one with some level of normative intervention (mitigation). Fur-thermore, a distinction is drawn here between scenarios and emissions pathways.Whereas the latter focus solely on emissions, a scenario represents a more completedescription of possible future states of the world, including their socio-economiccharacteristics and energy and transport infrastructures. Under this definition, manyof the existing ‘scenarios’ are in fact pathways, including the ones derived in thisstudy. Following the distinction between ‘emission scenarios’ and ‘concentrationprofiles’ introduced by Enting et al. (1994), the term ‘profiles’ is here used for timetrajectories of concentrations.

Existing mitigation pathways or scenarios differ in many respects, for examplein regard to the type and level of their envisaged climate targets (see overview inTable I).

One of the major challenges for the design of global mitigation pathways is thebalanced treatment of CO2 and non-CO2 emissions over a range of climate targetswith varying levels of stringency. Another major challenge is highlighted by the

MULTI-GAS PATHWAYS 153

TABLE IOverview of intervention pathways and scenarios

Name Climate Target Characteristic/comment Reference

‘S’ profiles CO2 concentrationstabilization at350, 450, 550,650 and750 ppm

CO2 profiles developed as partof a carbon-cycleinter-comparison exercise(Enting et al., 1994). CO2

emissions departed from‘business-as-usual’ in 1990.CO2 emissions varied only.

(Enting et al., 1994;Houghton et al.,1994)15

‘WRE’profiles

CO2 concentrationstabilization at350, 450, 550,650, 750 and1000 ppm

Variant of ‘S’ profiles with alater departure from‘business-as-usual’ emissionsdepending on the targetconcentration level. CO2

emissions varied only.

(Wigley et al., 1996)

Post-SRES(IPCCstabilizationscenarios)

CO2 concentrationstabilization atlevels between440 and 750ppm

Emission scenarios developedduring and subsequent to thework for the Special Reporton Emission Scenarios(SRES) (Nakicenovic andSwart, 2000). Modeldependent coverage andvariation of major greenhousegases and other radiativeforcing agents.

Different modelinggroups, namely AIM,ASF, IMAGE, LDNE,MARIA, MESSAGE,MiniCAM, PETRO,WorldScan (see e.g.Morita et al., 2000;and figure 2-1d inNakicenovic andSwart, 2000)16

TGCIA450 CO2 concentrationstabilization at450 ppm

Single pathway with coverageof all major greenhouse gasesand radiative forcing agentsto complement thenon-intervention SRESillustrative scenarios forAOGCM based climateimpact studies.

(Swart et al., 2002)

IMAGE Se CO2 equivalentconcentrationstabilization(based onradiative forcingof all GHGsincluded inKyoto Protocol)at 550, 650 and750 ppm

Following the concept ofstabilizing CO2 equivalentconcentrations (Schimelet al., 1997), the IMAGEteam designedCO2-equivalent emissionspathways (based on 100-yearGWP) with both (a) non-CO2

GHG emissions leading to100 ppm CO2 equivalentconcentrations and (b)non-CO2 emissions accordingto cost-optimal mixes.

(Eickhout et al., 2003)(van Vuuren et al.,2003)

(Continued on next page)


TABLE I(Continued)

Name Climate target Characteristic/comment Reference

MESSAGE-WBGU’03

CO2 concentrationstabilization at400 and450 ppm

Three intervention scenariosgenerated with MESSAGE forenergy-related CO2 andnon-CO2 emissions based ondifferent SRES baselines(A1-450; B1-400; B2-400).Non-energy related emissionsbased on AIM model.Commissioned by WBGU(2003).

(Nakicenovic and Riahi,2003)

EMF-21 Radiative forcingstabilization at4.5 W/m2

Baseline and model-dependent,cost-optimized scenarios forall major greenhouse gasesand other radiative forcingagents. To be published.

Various modelinggroups; (de laChesnaye, 2003)

EQW Freely selectable17 Emissions pathways with allmajor radiative forcing agents‘consistently’ varying with thestringency of climate target.Freely selectable departureyear from ‘business-as-usual’.

This study

debate on ‘early action’ versus ‘delayed response’ (see e.g. Ha-Duong et al., 1997;see e.g. Azar, 1998). Both issues arise from the fact that a long-term concentration,temperature or sea-level target can be achieved through more than one emissionspathway. Emissions in one gas (e.g. CO2) can be balanced against reductions inanother gas (e.g. N2O), which leads to a ‘multi-gas indeterminacy’. This is some-what parallel to the debate on the ‘timing of emission reductions’, since emissionsin the near-term may be balanced against reductions in the long-term. Obviously,there is a clear difference too: The ‘timing’ of emission reductions touches inter-generational equity questions much more directly than trade-offs between gases.Only indirectly, trade-offs between gases might have some implications for in-tergenerational issues, e.g. if states operate under a ‘Global Warming Potential’(GWP) based commitment period regime for gases of different lifetimes (Smithand Wigley, 2000b; Sygna et al., 2002). This paper proposes a method, which ischaracterized by its unique way of handling the ‘multi-gas indeterminacy’.

In the next section we review previous approaches to handling non-CO2 gasesin intervention pathways and in climate impact studies (Section 2). The ‘EqualQuantile Walk’ (EQW) method is presented subsequently (Section 3). EQW gen-erated multi-gas pathways are presented and compared with existing mitigationpathways (Section 4). Limitations of the EQW method are discussed subsequently


(Section 5). Finally, we conclude and suggest future work that can build on thepresented method (Section 6).

2. Previous Approaches to Handling Non-CO2 Gases in Mitigation

Pathways and Climate Impact Studies

To date, four different approaches have been used to handle the treatment of non-CO2 emissions in mitigation pathways. The simplest and most widely applied ap-proach we term here the ‘one size fits all’ approach, which means that different CO2

pathways are complemented by a single set of non-CO2 emissions. For example,the IPCC Second Assessment Report (SAR) focused only on CO2 when assessingstabilization scenarios (see IPCC, 1996, section 6.3). The temperature implicationsof the S profiles (see Table I) were thus derived in the SAR by assuming constantemissions for SO2 and constant concentrations for non-CO2 greenhouse gases attheir 1990 levels. Subsequently, Schimel et al. (1997) presented estimates of hownon-CO2 emissions might change in the future for the S profiles. Azar and Rhode(1997) presented temperature implications of the S profiles by assuming a 1W/m2

contribution by other greenhouse gases and aerosols. However, the non-CO2 emis-sions or radiative forcing contributions were still assumed to be independent of theCO2 stabilization levels. The Third Assessment Report (IPCC TAR) presented thetemperature effects of S and WRE profiles by assuming a common non-interventionscenario (SRES A1B) for non-CO2 emissions (see figure 9.16 in Cubasch et al.,2001).

Clearly, it is inconsistent to assume ‘non-intervention’ scenarios for non-CO2

gases in a general ‘climate-policy’ intervention scenario. An overestimation of theassociated effect on global mean temperatures for a certain CO2 concentration islikely to be the result. There are a number of ways in which non-CO2 gases mightbe accounted for more realistically, including the approach presented in this paper.Mitigation scenarios might want to assume a consistent mix of climate and air pol-lution related policy measures to lower CO2 emissions as well as to make use ofthe extensive non-CO2 mitigation potentials (see e.g. de Jager et al., 2001). Fur-thermore, constraints on carbon emissions are likely to be automatically correlatedwith lower non-CO2 emissions from common sources (e.g. limiting the burningof fossil fuels generally results in both, lower CO2 and lower aerosol emissions).Indeed, the approaches described below take account of such correlations betweenCO2 and non-CO2 gases in various ways.

The second approach that has been used may be referred to as ‘scaling’ and wasfirst employed by Wigley (1991). Non-CO2 emissions, concentrations or radiativeforcing are proportionally scaled with CO2. Some studies analysed the S profilesand accounted for non-CO2 gases, including sulphate aerosols, by scaling the ra-diative forcing of CO2. For example, the combined cooling effect of SO2 aerosoland warming effect of non-CO2 greenhouse gases has been assumed to add 23%


to the CO2 related radiative forcing in Wigley (1995) and Raper et al. (1996); 23%is the 2100 average for the 1992 IPCC emission scenarios (Leggett et al., 1992)according to Wigley and Raper (1992). Later, aerosols and greenhouse gases havebeen treated separately. For both the S and WRE-profiles, SO2 emissions wereeither held constant at their 1990 levels or the negative forcing due to sulphateaerosols (‘S(x)’) was directly scaled with changes in CO2 emissions since 1990(‘F(x)/F(1990)’), according to S(x) = [S(1990)/F(1990)]∗F(x). The scaling proce-dure for sulphate emissions was a significant improvement to explicitly capture thecorrelated nature of SO2 and fossil CO2 emissions. The positive forcing of non-CO2

greenhouse gases has then been assumed to be 33% of the CO2 related radiativeforcing (Wigley et al., 1996).

A third approach is to take source-specific reduction potentials for all gases intoaccount. Thus, rather than assuming that proportional reductions are possible acrossall gases, emission scenarios are developed by making explicit assumptions aboutreductions of the different gases. Realized reductions vary with the stringency ofthe climate target. In case of most of the Post-SRES scenarios, reductions in non-CO2 emissions result from systemic changes in the energy system as a result ofpolicies that aim to reduce CO2 emissions. This in particular involves CH4 fromenergy production and transport (see e.g. Post-SRES scenarios as presented inMorita et al. (2000), and Swart et al. (2002)). This method does not directly takeinto account the relative costs of reductions for different gases.

A fourth, more sophisticated, approach is to find cost-optimizing mixes of gas-to-gas reductions with the help of more or less elaborated energy and land-usemodels. In its simplest form, a set of (time-dependent) Marginal Abatement Costcurves (MAC) for different gases are used, thus enabling the determination of anoptimal set of reductions across all gases (see e.g. den Elzen and Lucas, 2005).Some studies mix both model-inherent cost estimates and exogenous MACs (seee.g. van Vuuren et al., 2003; den Elzen et al., 2005). Ideally, dynamically coupled(macro-)economic-energy-landuse models could aim to find cost-effective reduc-tion strategies that take into account model-specific assumptions about endogenoustechnological development, institutional and regulatory barriers as well as otherdriving forces for CO2 and non-CO2 emissions. Some of the more sophisticatedmodels within the Energy Modelling Forum (EMF) 21 model-inter-comparisonstudy aim to do so (de la Chesnaye, 2003).

One important distinction among scenarios of this fourth ‘cost-optimizing’ ap-proach can be drawn in regard to what exactly the modeling groups optimize.Some optimizing methods handle the ‘multi-gas indeterminacy’ by finding a cost-optimizing solution for matching a prescribed aggregated emission path (see e.g.den Elzen and Meinshausen, 2005). In this way the substitution between gasesis done using GWPs, which closely reflects current political (emission trading)frameworks. A different method is to determine gas-to-gas ratios by finding a cost-efficient emission path over time to match a long-term climate target. In this latterapproach, GWPs are not used to determine the substitution between gases but an


intertemporal optimization is performed to find cost-efficient emission paths to-wards a certain climate target. In general, the outcomes of these two optimizationmethods can be rather different, with the GWP-based approaches suggesting earlierand deeper cuts of short-lived greenhouse gases. The latter intertemporal optimiza-tion approaches rather advise to solve the ‘multi-gas indeterminacy’ in favor ofreductions of long-lived gases from the beginning with reductions of short-livedgases, such as CH4, only becoming important closer to times, when the climatetarget might be overshoot.

Whilst the ‘one size fits all’ and the ‘scaling’ approaches have the virtue ofcomputational simplicity, they have the clear disadvantage that the emission levelsfrom the non-CO2 gases and forcing agents may be economically or technologically‘unrealistic’. In other words, the assumed contribution of non-CO2 gases and forc-ing agents is unlikely to be consistent with the underlying literature on multi-gasgreenhouse mitigation scenarios based, for example, on methods three and four.The much more sophisticated third and fourth methods described here have thecompelling advantage of generating multi-gas pathways consistent with a processbased understanding of emission sources and control options and their relationshipto other economic factors, as well as dynamic interactions amongst sectors–as inthe case of the more sophisticated studies within method four. These methods areusually based on integrated assessment models (e.g. MESSAGE, IMAGE, AIMetc). So far, the volume of output and the complexity of input assumptions andrelated databases has militated against their use for generating large numbers ofscenarios for arbitrary climate targets and different time paths of emissions. How-ever, a solid exploration of emission implications of climate targets would requiresensitivity studies with (large ensembles) of multi-gas mitigation pathways.

Thus, the EQW method offers a computationally flexible approach to derivemulti-gas emissions pathways for a wide range of climate targets and scientificparameters, by extending and building upon scenarios under approaches three andfour above. Obviously, EQW pathways are an amendment to, but not a replace-ment of the mitigation scenarios of approaches three and four. The generation ofEQW pathways vitally depends on such mitigation scenarios, which capture thecurrent knowledge on mitigation potentials. There are numerous questions thatare best answered by specific scenarios under approaches three and four, e.g. inregard to implications for energy infrastructure and economic costs, which can-not be answered by EQW emissions pathways alone. However, EQW pathwaysare a vital extension, when it comes to explore the (multi-gas) emission impli-cations under various kinds of climate targets, possibly in a probabilistic frame-work (see e.g. Section 4.2). Whether certain emission reductions are consideredfeasible is outside the scope of this study and is a judgment that is likely tochange over time as new insights into technological, institutional, managementand behavioral options are gained. Furthermore, the EQW pathways might be usedto append CO2-only scenarios with a corresponding set of non-CO2 emissionspathways.


Many climate impact studies that explore climate change mitigation futuresreflect the scarcity of fully developed multi-gas mitigation pathways to date. Forexample, Arnell et al. (2002) and Mitchell et al. (2000) made assumptions similarto those used in the IPCC SAR (IPCC, 1996, Section 6.3). Their implementationof the S750 and S550-profiles assumes constant concentrations of non-CO2 gasesat 1990 levels, but did not consider forcing due to sulphate aerosols. Some studiesbound CO2 concentrations at a certain level, e.g. 2× or 3× pre-industrials levels,after having followed a ‘no climate policy’ reference scenario, e.g. IS92a (see e.g.Cai et al., 2003). Other studies assume ‘no climate policy’ trajectories for non-CO2 gases, thereby focusing solely on the effect of CO2 stabilization (Dai et al.,2001a,b) – although it should be noted that theses studies made a deliberate choiceto consider the effects of CO2 reductions alone in order to explore sensitivities in acontrolled way.

3. The ‘Equal Quantile Walk’ Method

We will refer to the presented method as the ‘Equal Quantile Walk’ (EQW) ap-proach for reasons explained below. A concise overview on the consecutive stepsof the EQW method is provided in Figure 1. The approach aims to distil a ‘distri-bution of possible emission levels’ for each gas, each region and each year out ofa compilation of existing non-intervention and intervention scenarios in the litera-ture that use methods three and four above (see Figure 1 and Section 2). Once thisdistribution is derived, which is notably not a probability distribution (cf. Section5.1.2), emissions pathways can be found, that are ‘comparably low’ or ‘comparablyhigh’ for each gas. In this way the EQW method builds on the sophistication anddetailed approaches that are inherent in existing intervention and non-interventionscenarios without making its own specific assumptions on different gases’ reductionpotentials.

Here, the term ‘comparably low’ is defined as a set of emissions that are on thesame ‘quantile’ of their respective gas and region specific distributions. Hence, theapproach is called ‘Equal Quantile Walk’ (cf. Figure 3 and Section 3.3). For exam-ple, the quantile path can, over time, be derived by prescribing one specific gas’semissions path in a particular region, such as fossil fuel CO2 for the OECD region(Section 3.2). The corresponding quantile path is then applied to all remaininggases and regions and a global emissions pathway is obtained by aggregating overthe world regions (Section 3.3). Consequently, EQW pathway emissions for one gascan go up over time, while emission of another gas go down, but an EQW pathwayfor a more ambitious climate target will be assumed to have lower emissions acrossall gases compared to an EQW pathway for a less ambitious climate target. Subse-quently, a simple climate model is used to find the corresponding profiles of globalmean temperatures, sea levels and other climate indicators. Here we use the simpleclimate model MAGICC 4.1 (Wigley and Raper, 2001, 2002; Wigley, 2003a). This


I

II-x%

-y%

OECD REF ASIA ALM

Simple Climate Model MAGICC

SRES /Post-SRES

Target achieved?

-

CO2CO2CO2CO2

CO2N2O

CO2CO2CO2CO2

CO2N2O

CH4

CO2

+

SiMCaP - Pathfinder

(1)

I

0

0.5

1(2) (3)

(4)(5)

CO2CO2CO2CO2

CO2N2O

CO2CO2CO2CO2

CO2N2O

CH4

CO2

Emission pathway

Climate target

Output emission pathway

CO2CO2CO2CO2

CO2N2O

CO2CO2CO2CO2

CO2N2O

CH4

CO2

CO2CO2CO2CO2

CO2N2O

CO2CO2CO2CO2

CO2N2O

CH4

CO2

CO2CO2CO2CO2

CO2N2O

CO2CO2CO2CO2

CO2N2O

CH4

CO2

CO2CO2CO2CO2

CO2N2O

CO2CO2CO2CO2

CO2N2O

CH4

CO2

OEC

D fo

ssil

CO

2 Driver path Quantile Path

Climate output

Distributions of possible emission levels for each year t

t t

tt

t

t

t

t

t

t

Figure 1. The EQW method as implemented in SiMCaP’s ‘pathfinder’ module. (1) The ‘distributionsof possible emission levels’ are distilled from a pool of existing scenarios for the 4 SRES worldregions OECD, REF, ASIA and ALM.13 (2) The common quantile path of the new emissions pathwayis derived by using a driver emission path, such as the one for fossil CO2 emissions in OECD countries.The driver path is here defined by sections of constant emission reductions (‘−x/y%’) and years atwhich the reduction rates change (‘I’ and ‘II’). (3) A global emissions pathway is obtained by assumingthat – in the default case – the quantile path that corresponds to the driver path applies to all gasesand regions. (4) Using the simple climate model MAGICC, the climate implications of the emissionspathway are computed. (5) Within SiMCaP’s iterative optimisation procedure, the quantile paths areoptimised until the climate outputs and the prescribed climate target match sufficiently well.

is the model that was used for global-mean temperature and sea level projectionsin the IPCC TAR (see Cubasch et al., 2001 and Section 3.4 and Appendix A).

An iterative procedure is used to find emissions pathways that correspond to apredefined arbitrary climate target. This is implemented in the ‘EQW pathfinder’module of the ‘Simple Model for Climate Policy Assessment’ (SiMCaP). Morespecifically, SiMCaP’s iterative procedure begins with a single ‘driver’ emissionpath (such as fossil CO2 in the OECD region) and then uses the ‘equal quantile’assumption to define emissions for all other gases and regions. The driver pathis then varied until the specified climate target is sufficiently well approximatedusing a least-squares goodness of fit indicator (see Figure 1). SiMCaP’s model


components and a set of derived EQW emissions pathways are available from theauthors or at http://www.simcap.org.

3.1. DISTILLING A DISTRIBUTION OF POSSIBLE EMISSION LEVELS

In order to determine a possible range of different gases’ emission levels a set ofscenarios is needed. Here, the 40 non-intervention IPCC emission scenarios fromthe Special Report on Emission Scenarios (Nakicenovic and Swart, 2000)2 areused in combination with 14 Post-SRES stabilization scenarios from the same sixmodeling groups,3 as presented by Swart et al. (2002). This combined set of 54scenarios is used in this study to derive the distributions of possible emission lev-els. The Post-SRES intervention scenarios are scenarios that stabilize atmosphericCO2 concentrations at levels between 450 ppm to 750 ppm. Most of the Post-SRES scenarios only target fossil CO2 explicitly, although lower non-CO2 emis-sions are often implied due to induced changes on all energy-related emissions.For halocarbons (CFCs, HCFCs and HFCs) and other halogenated compounds(PFCs, SF6), the post-SRES scenarios, however, provide no additional informa-tion. Therefore, the A1, A2, B1 and B2 non-intervention IPCC SRES scenarioswere supplemented with one intervention pathway in order to derive the distribu-tion of possible emission levels. Since most of the halocarbons and halogenatedcompounds can be reduced at comparatively low costs compared to other gases (cf.USEPA, 2003; Ottinger-Schaefer et al., submitted), the added intervention path-way assumes a smooth phase-out by 2075. Clearly, future applications of the EQWmethod can be based on an extended set of underlying multi-gas scenarios (such asEMF-21), thereby capturing the best available knowledge on multi-gas mitigationpotentials.

The combined density distribution for the emission levels of the different gaseshas been derived by assuming a Gaussian smoothing window (kernel) around eachof the 54 scenarios. The resulting non-parametric density distribution for a givenyear and gas can be viewed as a smoothed histogram of the data (see Figure 2). Anarrow kernel would reveal higher details of the underlying data until every singlescenario is portrayed as a spike–as in a high-resolution histogram. Wider kernels canalso be used to some degree to interpolate and extrapolate information of the limitedset of reduction scenarios into underrepresented areas within and outside the rangeof the scenarios. Thus, the chosen kernel width has to strike a balance between-onthe one hand-allowing a smooth continuum of emission levels and the design ofslightly lower emissions pathways and – on the other hand – appropriately reflect-ing the lower bound as well as the possibly asymmetric nature of the underlyingdata.

In this study, a medium width of the kernel is chosen-close to the optimumfor estimating normal distributions (Bowman and Azzalini, 1997). For a limitednumber of cases a narrower kernel width was chosen, namely for the N2O related


0 1 2 3 4 5 6 7 (GtC)

0 50 100 150 200 250 (MtCH4)

a) OECD; Fossil CO2; 2050

b) OECD; CH4; 2100

narrow

narrow

medium

medium

wide

wide

den

sity

den

sity

Figure 2. Derived non-parametric density distribution by applying smoothing kernels with defaultkernel width for this study (solid line ‘medium’), a wide kernel width (dashed line ‘wide’) and anarrow kernel width (dotted line ‘narrow’). See text for discussion.

distributions in order to better reflect the lower bound of the distribution. A narrowerkernel for N2O guarantees a more appropriate reflection of the sharp lower boundof the distribution of N2O emission levels, which is suggested by the pool ofexisting SRES & post-SRES scenarios (see Figure 9c). The application of a widerkernel would have resulted in an extensive lapping of the derived non-parametricdistribution into low emission levels that are not represented within the set ofexisting scenarios. The inclusion of a wider set of currently developed multi-gasscenarios might actually soften this seemingly hard lower bound for N2O emissionsin the future.4 Furthermore, the distribution of possible emission levels might extendinto negative areas, which is, for most emissions, an implausible or impossiblecharacteristic. Thus, derived distributions have been truncated at zero with theexception of land-use related net CO2 emissions.

Land use CO2 emissions, or rather CO2 removal, have been bound at the lowerend according to the SRES scenario database literature range as presented in fig-ure SPM-2 of the Special Report on Emission Scenarios (Nakicenovic and Swart,2000). Specifically, the applied lower bound ranges between −1.1 and −0.6 GtC/yrbetween 2020 and 2100. The maximum total uptake of carbon in the terrestrialbiosphere from policies in this area over the coming centuries is assumed to ap-proximately restore the total amount of carbon lost from the terrestrial biosphere.Specifically, it was assumed that from 2100 to 2200, the lower bound for the land-userelated CO2 emission distribution smoothly returns to zero so that the accumulatedsequestration since 1990 does not exceed the deforestation related emissions be-tween 1850 and 1989, estimated to be 132 GtC5 (Houghton, 1999; Houghton andHackler, 2002).


3.2. DERIVING THE QUANTILE PATH

For an EQW pathway, emissions of each gas in a given year and for a given regionare assumed to correspond to the same quantile of the respective (gas-, year- andregion-specific) distribution of possible emission levels. Depending on the climatetarget and the timing of emission reductions, the annual quantiles might of coursechange over time (cf. inset (2) in Figure 1). It is possible to prescribe the quantilepath directly. For example, aggregating emissions that correspond to the time-constant 50% quantile path would result in the median pathway over the wholescenario data pool. In general, however, what we do is prescribe one of the gases’emissions as ‘driver path’, for example the one for fossil CO2 emissions in OECDcountries. The corresponding quantile path can then be applied to all other gasesin that region. If desired, the same quantile path may be applied to all regions. Fora discussion on the validity of such an assumption of ‘equal quantiles’ the readeris referred to Section 5.1.1 with alternatives being briefly discussed in Section5.1.7. Theoretically, one could for example also prescribe aggregate emissions asthey are controlled under the Kyoto Protocol (Kyoto gases) and any consecutivetreaties using 100-yr GWPs.14 Specifically, one could derive the correspondingquantile path by projecting the prescribed aggregate emissions onto the distributionof possible aggregate emission levels implied by the underlying scenarios. Suchquantile paths, possibly regionally differentiated due to different commitments,could then be applied to all gases individually in the respective regions, provided apool of standardized scenarios for the same regional disaggregation existed.

In this study, we have adopted a fairly conventional set of climate policy as-sumptions to derive the emissions pathways. One of the key agreed principles inthe almost universally ratified United Nations Framework Convention on ClimateChange (UNFCCC, Article 3.1) is that of “common but differentiated responsi-bilities and respective capabilities” which requires that “developed country Partiesshould take the lead in combating climate change”.1 As a consequence, it is ap-propriate to allow the emission reductions in non-Annex I regions6 to lag behindthe driver. Furthermore, a constant reduction rate (exponential decline) of absoluteOECD fossil CO2 emissions has been assumed for ‘peaking’ scenarios after a pre-defined ‘departure year’ from the baseline emission scenario (here assumed to bethe median over all 54 IPCC scenarios). For ‘stabilization’ scenarios, the annualrate of reduction was allowed to change once in the future in order to lead to thedesired stabilization level (see inset 2 within Figure 1). A constant annual emissionreduction rate has been chosen for two reasons: (a) simplicity, and (b) because ofthe fact that such a path is among those that minimize the maximum of annualreductions rates needed to reach a certain climate target.

Up to the predefined departure year, e.g. 2010, emissions follow the medianscenario (quantile 0.5; cf. Figure 3). The departure year can differ from region toregion and indeed, as noted above, this is required by the UNFCCC and codifiedfurther in the principles, structure and specific obligations in the Kyoto Protocol.


0 0.5 1100

101

102

103

104

OEC

D E

mis

sio

ns

(MtC

O2e

q)

0 0.5 1 0 0.5 1

Fossil CO2

CH4

N2O

CF4

C2F6

HFC-134a

SF6

QuantileQuantile Quantile

Fossil CO2

CH4

N2O

CF4

C2F6

HFC-125

HFC-134a

HFC-143a

HFC-227eaHFC-245ca

SF6

Fossil CO2

CH4

N2O

CF4

C2F6

HFC-125

HFC-134a

HFC-143a

HFC-227eaHFC-245ca

SF6

Year 2000 Year 2050 Year 2100

Figure 3. The derived distributions of possible emission levels displayed as (inverse) cumulativedistribution functions for OECD countries in the years 2000 (right), 2050 (middle) and 2100 (left).The nearly horizontal lines for the year 2000 (left panel) illustrate that all 54 underlying scenariosshare approximately the same emission level assumptions for the year 2000 (basically because thesescenarios are standardized). In later years, here shown for 2050 and 2100, the scenarios’ projectedemissions diverge, so that the lower percentile (left side of each panel) corresponds to lower emissionscompared to the upper percentile (right side of each panel) of the emission distributions. Thus, theslope of cumulative emission distribution curves goes from lower-left to upper-right. New mitigationpathways are now constructed by assuming a set of emissions for each year that corresponds to thesame quantile (black triangles) in a respective year. These quantiles can for example be chosen so thata prescribed emission path for fossil CO2 is matched. The non-fossil CO2 emissions are then chosenaccording to the same quantile (see dots on dashed vertical lines). The same procedure is appliedto other non-OECD world regions by using either the same or different quantile path (see text). Forthis illustrative figure (but not for any of the underlying calculations within the EQW method), allemissions have been converted to Mt CO2-equivalent using 100-yr GWPs.14 Note the logarithmicvertical scale, which causes zero and negative emissions not being displayed.

Here non-Annex I countries are assumed to diverge from the baseline scenarioa bit later (2015) than Annex-I countries (2010) and follow a quantile path thatcorresponds to a hypothetically delayed departure of fossil fuel CO2 emissions inOECD countries.

Generally, it should be noted that there could be a difference between actualemissions and the assumed emission limitations in each region to the extent thatemissions are traded between developed and developing countries.


3.3. FINDING EMISSIONS PATHWAYS

Once the non-parametric distributions of possible emission levels (Section 3.1) aredefined and the quantile paths (3.2) prescribed, multi-gas emissions pathways forany possible climate target can be derived. For any specific year, the emission levelsof each greenhouse gas and aerosol for different regions are selected according toa specific single quantile for the particular year and region. This will result in aset of emissions that is ‘comparably low’ or ‘high’ in relation to the underlyingpool of existing emission scenarios (see Figure 3). As a final step a smoothingspline algorithm has been applied to the individual gases pathways other than thedriver path, restricted to the years after the regions’ departure year from the baselinescenario.

3.4. THE CLIMATE MODEL

All major greenhouse gases and aerosols are inputs into the climate model, namelycarbon dioxide (CO2), methane (CH4), nitrous oxide (N2O), the two most relevantperfluorocarbons (CF4, C2F6), and five most relevant hydrofluorocarbons (HFC-125, HFC-134a, HFC-143a, HFC-227ea, HFC-245ca), sulphur hexafluoride (SF6),sulphate aerosols (SO2), nitrogen oxides (NOx ), non-methane volatile organic com-pounds (VOC), and carbon monoxide (CO). Emissions of these gases are calculatedfor the different climate targets using the EQW method. Thus, these emissions werevaried according to the stringency of the climate target. For the limited number of re-maining human-induced forcing agents, the assumed emissions follow either a ‘onesize fits all’ or ‘scaling’ approach, due to the lack of data within the pool of SRESand Post-SRES scenarios. Specifically, the forcing due to substances controlledby the Montreal Protocol is assumed to be the same for all emissions pathways.Similarly, emissions of other halocarbons and halogenated compounds aside fromthose eight explicitly modeled are assumed to return linearly to zero over 2100 to2200 (‘one size fits all’). The combined forcing due to fossil organic carbon andblack organic carbon was scaled with SO2 emissions after 1990 (‘scaling’), as inthe IPCC TAR global-mean temperature calculations.

A brief description of the default assumptions made in regard to the employedsimple climate model MAGICC and natural forcings are given in the Appendix.

4. ‘Equal Quantile Walk’ Emissions Pathways

The following section presents some results in order to highlight some of the keycharacteristics of the EQW method. First, we compare the results of the EQWmethod with previous CO2 concentration stabilization pathways. It is shown thatthere can be a considerable difference in terms of non-CO2 forcing for the same CO2


stabilization level, which is the result of EQW pathways taking into account thenon-CO2 mitigation potentials to the extent that they are included in the underlyingmulti-gas scenarios. Second, we examine two sets of peaking pathways, wherethe global mean radiative forcing peaks and hence where concentrations do notnecessarily stabilize (not as soon as under CO2 stabilization profiles at least). Inprinciple, these may be useful in examining emissions pathways correspondingto climate policy targets that recognize that it may be necessary to lower peaktemperatures in the long term in order to take account of–for example–concernsover ice sheet stability (Oppenheimer, 1998; Hansen, 2003; Oppenheimer and Alley,2004). Provided one makes specific assumptions on the most important climateparameters, such as climate sensitivity, one could also derive temperature (rate)limited pathways (not shown in this study).

4.1. COMPARISON WITH PREVIOUS PATHWAYS

This section compares EQW multi-gas emissions pathways with emissions of the Sand WRE CO2 stabilization profiles. In order to allow a comparison between theseemissions pathways, sample ‘EQW’ emissions pathways were designed to reachCO2 stabilization at 350 to 750 ppm. After the default departure years (2010 forAnnex I regions and 2015 for non-Annex I), the quantile path corresponds to a rateof reduction of OECD fossil CO2 emissions between −5.2% and −0.5% annuallydepending on the stabilization level. These annual emission reductions are adjustedat a point in the future (derived in the optimization procedure) in order to allowCO2 concentrations to stay at the prescribed stabilization levels (see Table II).

While fossil CO2 emissions between WRE and these sample EQW pathwaysconverge in the long-term, the near and medium-term fossil CO2 emissions differ(see Figure 4). For the lower stabilization levels, the assumptions chosen here forthe EQW pathways lead to slightly higher fossil CO2 emissions than the WREpathways, which is mainly due to the fact that the land-use related CO2 emissionsare substantially lower under the EQW than under WRE. For the same reason,cumulative fossil CO2 emissions are slightly higher for the EQW pathways thanfor the corresponding WRE pathways (not shown in figures). For the less stringentprofiles, namely stabilization levels between 550 and 750 ppm, the EQW assump-tions lead to fossil CO2 emissions that are lower in the near term, but decline moreslowly and are higher in the 22nd century and beyond. The main reason for thisdifference might be of a methodological nature rather than founded on differingexplicit assumptions on ‘early action’ vs. ‘delayed response’. As for the originalS profiles and many recent stabilization profiles (Eickhout et al., 2003), the WREprofiles were defined as smoothly varying CO2 concentration curves using Padeapproximants (cf. Enting et al., 1994) and emissions were determined by inversecalculations. In contrast to this ‘top-down’ approach, the EQW method can be cat-egorized as a ‘bottom-up’ approach in the transient period up to CO2 stabilization,


TAB

LE

IIR

educ

tion

rate

sfo

rO

EC

Dfo

ssil

CO

2(d

rive

rpa

ths)

,W

orld

foss

ilC

O2

and

Wor

ldag

greg

ated

Kyo

toga

ses

(6-G

HG

s)w

ithan

dw

ithou

t‘O

ther

CO

2’

that

are

com

patib

lew

ithre

achi

ngC

O2

stab

ilisa

tion

leve

lsfr

om35

0to

750

ppm

acco

rdin

gto

the

EQ

Wm

etho

d.A

fter

the

depa

rtur

eye

ars

(201

0/20

15fo

rA

nnex

-I/n

on-A

nnex

I),O

EC

Dfo

ssil

CO

2em

issi

ons

are

assu

med

tode

crea

seat

aco

nsta

ntra

te(‘

Red

uctio

nra

teI’

).Fr

omth

e‘a

djus

tmen

tye

ar’

onw

ards

,th

ean

nual

emis

sion

redu

ctio

nra

teis

redu

ced

inor

der

tost

abili

zeC

O2

conc

entr

atio

ns(‘

Red

uctio

nra

teII

’).T

heth

ree

pres

ente

dpa

ram

eter

valu

es,r

educ

tion

rate

Ian

dII

and

the

adju

stm

enty

ear,

are

optim

alin

the

sens

e,th

atth

ere

sulti

ngC

O2

conc

entr

atio

npr

ofile

sbe

stm

atch

the

pres

crib

edst

abili

zatio

nle

vels

unde

ra

leas

t-sq

uare

sgo

odne

ss-o

f-fit

indi

cato

r.O

ther

sour

ces’

and

gase

s’em

issi

ons

follo

wth

esa

me

quan

tile

path

(see

Sect

ion

3.3)

resu

lting

inva

riab

lew

orld

wid

ere

duct

ion

rate

sfo

rfo

ssil

CO

2an

dag

greg

ated

emis

sion

s(u

sing

100-

yrG

WPs

)ov

ertim

e.

Wor

ldFo

ssil

Wor

ld6-

GH

Gs

Wor

ld6-

GH

Gs

CO

2ex

cl.‘

Oth

erC

O2’

incl

.‘O

ther

CO

2’

OE

CD

Foss

ilC

O2

CO

2R

ange

redu

ctio

nR

ange

redu

ctio

nR

ange

redu

ctio

nst

abili

zatio

nR

educ

tion

rate

Adj

ustm

ent

Red

uctio

nra

tera

tes

2020

–210

0ra

tes

2020

–210

0ra

tes

2020

–210

0le

vel(

ppm

)I

(%/y

ear)

year

II(%

/yea

r)(%

/yea

r)(%

/yea

r)(%

/yea

r)

350

−5.1

721

200.

00−1

.05

to−4

.64

−0.6

1to

−3.2

4−0

.28

to−5

.19

450

−2.1

820

70−0

.74

−0.6

2to

−1.3

1−0

.65

to−0

.93

−0.7

1to

−1.5

7

550

−0.9

322

11−0

.38

+0.6

2to

−1.0

1+0

.40

to−0

.82

+0.0

3to

−1.1

2

650

−0.6

323

27−0

.01

+0.8

6to

−0.6

7+0

.66

to−0

.59

+0.4

0to

−0.7

7

750

−0.4

623

790.

00+0

.98

to−0

.48

+0.8

0to

−0.4

4+0

.59

to−0

.55


PIL

+1

+2

+3

+4

+5

2000 2100 2200 2300 2400-2

0

2

4

6

8

10

12

14

2000 2100 2200 2300 2400PIL

350

400

450

500

550

600

650

700

750

800b) CO2 Concentrations (ppmv)

c) Temperature (˚C)

a) Fossil CO2 Emissions (GtC)

d) Sea Level (cm)

WRE

2000 2100 2200 2300 24002000 2100 2200 2300 2400PIL

+20

+40

+60

+80

+100

+120

750

650

550

350

450

750

650

550

350

450

750

650

550

350

450

750

650

550

350

450

EQW

Figure 4. Comparison of WRE profiles (dashed) with EQW profiles (solid). (a) Fossil CO2 pathwaysdiffer (see text) for the (b) prescribed CO2 stabilization levels at 350, 450, 550, 650 and 750 ppm.(c) Global mean surface temperature increases above pre-industrial levels (‘PIL’) are lower for theEQW profiles for any CO2 stabilization (c) due to lower non-CO2 emissions. Correspondingly, sealevel increases are lower for EQW profiles (d). As in the IPCC TAR (cf. figure 9.16 in Cubasch et al.,2001), the WRE CO2 emissions pathways are here combined with non-CO2 emissions according tothe IPCC SRES A1B-AIM scenario (dashed lines).

since the profile towards CO2 stabilization is prescribed by multiple constraints onthe fossil CO2 emissions rather than on CO2 concentrations themselves.

Under the most stringent of the analyzed CO2 concentration targets, stabilizationat 350 ppm, near term fossil CO2 emissions depart, slightly delayed, from thebaseline scenario in comparison to the WRE350 pathway, which assumes a globaldeparture in 2000 (cf. Figure 4). Compared to the S-profiles, this difference (forall stabilization levels) is even larger as the S profiles assume an early start of


emission reductions in the 1990s and a smoother path thereafter, which alreadyseems unachievable today, due to recent emissions increases.

A comparison including non-CO2 gases can be done using the WRE profilesas they are presented in the IPCC TAR (see figure 9.16 in Cubasch et al., 2001).There, the effects for the WRE CO2 stabilization profiles are computed by assumingnon-CO2 gas emissions according to the A1B-AIM scenario (see Figure 4 andFigure 5. For the comparison, it is thus important to keep in mind that the EQWpathways are not compared to the WRE CO2 profiles per se, but to the WREpathways in combination with this specific assumption for non-CO2 emissions.

The EQW method chooses non-CO2 emissions on the basis of the CO2 quantile,which for all analyzed CO2 stabilization profiles implies that it chooses emissionssignificantly below the A1B-AIM levels–as also the fossil CO2 emissions are belowthose of the A1B-AIM scenario. Mainly due to these lower non-CO2 emissions, theradiative forcing implications related to EQW pathways are significantly reducedfor the same CO2 stabilization level when compared to WRE pathways, i.e. forstabilization at 450 ppm (see Figure 5). Partially offsetting this ‘cooling’ effectis the reduced negative forcing due to decreased aerosol emissions. The negativeforcing from aerosols can be significant (cf. dark area below zero in Figure 5) andcan mask some positive forcing due to CO2 and other greenhouse gases. In theyear 2000, this masking is likely to be about equivalent to the forcing due to CO2

alone (the upper boundary of the “CO2” area is near the zero line in Figure 5).However, note that large uncertainties persist in regard to the direct and indirectradiative forcing of aerosols (see e.g. Anderson et al., 2003).7 The total radiativeforcing for the WRE450 scenario in 2400 is ca. 3.9 W/m2 and around 3 W/m2 forthe EQW-S450C.

Owing to the effect on radiative forcing, the lowered non-CO2 emissions thatresult from the EQW method lead to less pronounced global mean temperatureincreases in comparison to the WRE CO2 stabilization profiles in combinationwith the A1B-AIM non-CO2 emissions. For the same CO2 stabilisation levels, thecorresponding temperatures are about 0.5 ◦C cooler by the year 2400 (assuminga climate sensitivity of approximately 2.8 ◦C by computing the ensemble meanover 7 AOGCMs-see Appendix A). Consequently, the sea level rise is also slightlyreduced when assuming the EQW pathways (cf. Figure 4).

4.2. RADIATIVE FORCING (CO2 EQUIVALENT) PEAKING PROFILES

A variety of climate targets can be chosen to derive emissions pathways withthe EQW method. In this section, two sets of multi-gas emissions pathways arechosen so that the corresponding radiative forcing peaks between approximately2.6 and 4.5 W/m2 with respect to pre-industrial levels. The CO2 equivalent peakingconcentrations are 475 to 650 ppm (see Figure 6). No time-constraint is placed onthe attainment of the peak forcing.


1700 1800 1900 2000 2100 2200 2300 2400-2

-1

0

1

2

3

4

SO4DIR

SO4IND

BIOAER

CO2

CH4N2O

HALOtotTROPOZ

FOC+FBC

EQW-S450C

-1

0

1

2

3

4

5

Rad

iati

ve F

orc

ing

(W

/m2 )

Rad

iati

ve F

orc

ing

(W

/m2 )

SO4DIR

SO4IND

BIOAER

CO2

CH4

N2O

HALOtot

TROPOZ

FOC+FBC

WRE450(with non-CO2 acc. to SRES A1B-AIM)

Figure 5. Aggregated radiative forcing as a result of the WRE emissions pathway (upper graph) andthe EQW pathway (lower graph) for stabilization of CO2 concentrations at 450 ppm. Since the ‘EQW’multi-gas pathways take into account reductions of non-CO2 gases, the positive radiative forcing dueto CH4, N2O, tropospheric ozone (‘TROPOZ’), halocarbons and other halogenated compounds minusthe cooling effect due to stratospheric ozone depletion (‘HALOtot’) as well as the negative radiativeforcing due to sulphate aerosols (indirect ‘SO4IND’ and direct ‘SO4DIR’) and biomass burningrelated aerosols (‘BIOAER’) is substantially reduced. The combined warming and cooling due tofossil fuel related organic & black carbon emissions (‘FOC+FBC’) is scaled towards SO2 emissions(see text).

The first set ‘A’ of derived EQW peaking pathways assumes a fixed departureyear, but variable rates of emission reductions thereafter. The second set ‘B’ holdsthe reduction rates of the driver emission path constant, but assumes varying de-parture years. Specifically, the peaking pathways ‘A’ assume a departure from themedian emission scenario in 2010 for Annex I countries (IPCC SRES regions REF


& OECD13 and a departure in 2015 for non-Annex I countries (ASIA & ALM).OECD fossil CO2 emissions, the driver emission path, are assumed to decline ata constant rate, which differs between the individual pathways ‘A’, after the fixeddeparture year. The second set ‘B’ of peaking pathways assumes a departure yearfrom the median emission scenario between 2010 and 2050 for Annex I countries(5 years later for non-Annex I countries), and a 3% decline of OECD fossil CO2

driver path emissions. As highlighted in the method section, emissions in non-OECD regions and from non-fossil CO2 sources are assumed to follow the quantilepath corresponding to the preset driver path (see Figure 6, Section 3.2 and 3.3).

For the derived emissions pathways that peak between 470 and 555 ppm CO2eq,global fossil CO2 emissions are between 46% to 113 % of 1990 emission levels in2050 (see Table III) and 11% to 33% in 2100, depending on the peaking target.

In parallel to the greenhouse gas emissions, the EQW method derives aerosoland ozone precursor emissions that are ‘comparably low’ in regard to the underly-ing set of SRES/Post-SRES scenarios. Thus, despite the fact that sulphate aerosolprecursor emissions (SOx ) have a cooling effect, SOx emissions are assumed todecline sharply for the more stringent climate targets (see Table III). The linkagebetween SOx and CO2 emissions is also seen in mitigation scenarios from cou-pled socio-economic, technological model studies and is partially due to the factthat both stem from a common source, namely fossil fuel combustion (see as wellSection 5.1.1). Another reason is that mitigation scenarios represent future worldswhich inherently include environmental policies in both developed and developingcountries-where the abatement of acid deposition and local air pollution has usuallyeven higher priority than greenhouse gas abatement.

Depending on the shape of the emissions pathways (e.g. set A or B), and de-pending on the peak level between 470 and 650 ppm CO2eq, radiative forcing peaksbetween 2025 and 2100. After peaking, radiative forcing (CO2 equivalence concen-trations) stays significantly above pre-industrial levels for several centuries. This is

←−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−Figure 6. Two sets of multi-gas pathways derived with the EQW method. The two sets are distinct inso far as set A assumes a fixed departure year from the median emission path (2010) and a differentreduction rate thereafter (−7% to 0%) (A.1). The pathways of set B assume a fixed reduction rate forOECD fossil CO2 emissions (−3%/yr), but variable departure years. Emissions of other gases and inother regions follow the corresponding quantile paths (see text). For illustrative purposes, the GWP-weighted sum of greenhouse gas emission is shown in panels A.2 and B.2. Using a simple climatemodel, the radiative forcing implications of the multi-gas emissions pathways can be computed,here shown as CO2 equivalent concentrations with black dots indicating the peak values (A.3 andB.3). The temperature implications are computed probabilistically for each peaking pathway usinga range of different climate sensitivity pdfs (see text). In this way, one can illustrate the probabilityof overshooting a certain temperature threshold (here 2◦C above pre-industrial) under such peakingpathways given different climate sensitivity probability distributions (dashed lines in darker shadedarea of A.4 and B.4). Lighter shaded areas depict the probability of overshooting 2◦C in equilibriumin case that concentrations were stabilized and not decreased after the peak. The full set of emissiondata is available at http://www.simcap.org.


TABLE IIISpecifications (I), emission implications (II) and probabilities of overshooting 2◦C (III) forthree radiative forcing peaking pathways (cf. Figure 6). Departure years and annual OECDfossil CO2 emission reductions (‘driver path’) were prescribed. For illustrative purposes only,greenhouse gas emissions (CO2, CH4, N2O, HFCs, PFCs, and SF6) were aggregated using100-year GWPs14 including and excluding landuse related CO2 emissions (‘other CO2’). Themaximum CO2 equivalence concentration (radiative forcing) is shown and its associated prob-ability of overshooting 2◦C global mean temperature rise above pre-industrial for a range ofdifferent climate sensitivity probability density function estimates (see text). The probability ofovershooting is clearly lower for the peaking pathways, where concentrations drop after reach-ing the peak level, in comparison to hypothetical stabilization pathways, where concentrationsare stabilized at the peak.

Peaking Peaking Peakingpathway 1 pathway 2 pathway 3

I. Specifications

Set of pathway A A/B B

Departure years (Annex I/Non-Annex I) 2010/15 2010/15 2020/25

Driver path OECD fossil CO2 reduction −5%/yr −3%/yr −3%/yr

II. Emission implications

Emissions (1990 level) 2050 Emissions relative to 1990

Fossil CO2 (5.98 GtC) 46% 80% 113%

CH4 (309 Mt) 77% 94% 112%

N2O (6.67 TgN) 68% 76% 81%

GHG excl. other CO2 (8.72 GtCeq) 55% 82% 110%

GHG incl. other CO2 (9.82 GtCeq) 41% 65% 90%

SOx (70.88 TgS) 4% 13% 26%

III. Peak concentration and probability of overshooting

Peak concentration CO2eq ppm (radiative 470 (2.80) 503 (3.17) 555 (3.70)

forcing W/m2)18

Probability >2 ◦C (peaking) 5–60% 25–77% 48–96%

Probability >2 ◦C (stabilisation) 35–88% 49–96% 69–100%

mainly due to the slow redistribution processes for CO2 between the atmospheric,oceanic and abyssal sediment carbon pools.

The temperature response of the climate system is largely dependent upon its cli-mate sensitivity, which is rather uncertain. A range of recent studies have attemptedto quantify this uncertainty in terms of probability density functions (PDFs) (seee.g. Andronova and Schlesinger, 2001; Forest et al., 2002; Gregory et al., 2002;Knutti et al., 2003; Murphy et al., 2004). These studies are used here to computeeach emissions pathway’s probabilistic climate implications by running the simpleclimate model with an array of climate sensitivities, weighted by their respectiveprobabilities according to particular climate sensitivity PDFs. The probabilistic


temperature implications of the radiative forcing peaking pathway sets can then beshown in terms of their probability of overshooting a certain temperature threshold,here chosen as 2 ◦C above pre-industrial levels (see it Figure 6). The faster the radia-tive forcing drops to lower levels after the peak, the less time there is for the climatesystem to reach equilibrium warming. Thus, for peak levels of 550 ppm CO2eq andabove, the peaking pathways B involve slightly lower probabilities of overshootinga 2 ◦C temperature thresholds, as their concentrations decrease slightly faster thanfor the higher peaking pathways of set A. The probability of overshooting 2 ◦Cwould obviously be higher for both sets, if radiative forcing were not decreasingafter peaking, but stabilized at its peak value, as depicted by the lighter shadedareas in Figure 6 A.4 and B.4 (Azar and Rodhe, 1997; Hare and Meinshausen,2004; Meinshausen, 2005).

In summary, it has been shown that the EQW method can provide a useful toolto obtain a large numbers of multi-gas pathways to analyze research questions in aprobabilistic setting. Furthermore, the results suggest that if radiative forcing is notpeaked at or below 475 ppm CO2eq (∼2.8 W/m2) with declining concentrationsthereafter, it seems that an overshooting of 2 ◦C can not be excluded with reasonableconfidence levels (see Figure 6).

5. Discussion and Limitations

The following section discusses some of the potential limitations, namely thoserelated to the EQW method itself (Section 5.1), and those related to the underlyingpool of scenarios (Section 5.2). In addition, the use of a simple climate modelimplies some limitations briefly mentioned in Appendix A.

5.1. DISCUSSION OF AND POSSIBLE LIMITATIONS ARISING FROM THE

METHOD ITSELF

The following section briefly discusses several issues that are directly related to theproposed EQW method: namely the assumption of unity rank correlations (5.1.1);the question, whether the individual underlying scenarios are assumed to have acertain probability (5.1.2); regional emission outcomes (5.1.4); the baseline (in-)dependency (5.1.5); land-use change related emissions and their possible politicalinterpretations (5.1.5); alternative gas-to-gas and timing strategies (5.1.7); and theprobabilistic framework (5.1.8).

5.1.1. Unity Rank CorrelationNew emissions pathways produced with the EQW method will rank equally acrossall gases in a specific region for a specific year. In other words, an emissionspathway for a less stringent climate target (e.g. peaking at 550 ppm CO2eq) has


higher emissions for all gases and all regions compared to an emissions pathwayfor a less stringent climate target (e.g. 475 ppm CO2eq).

Note that this inbuilt unity rank correlation assumption of the EQW methoddoes not necessarily lead to positive absolute correlations between different gases’or regions’ emissions. In other words, for a particular EQW mitigation pathway,emissions of one gas, e.g. CO2 in Asia, might still be increasing in a particularyear, while emissions of another gas, e.g. methane in OECD, are already decreas-ing depending on the emission distributions in the underlying pool of emissionscenarios.

The unity rank correlation could be an advantage of the EQW approach. How-ever, it could also be a limitation in the presence of negative rank correlationsfor emissions: for example, if fossil fuel emissions were largely reduced due to areplacement with biomass, a negative correlation might arise between fossil fuelCO2 and biomass-burning related aerosol emissions, such as SOx , NOx etc. Thus,if fossil fuel CO2 emissions decrease, some aerosol emissions might increase. NOxand N2O emission changes may be negatively correlated up to a certain degree aswell. Coupled socio-economic, technological, and land use models, such as thoseused for creating the SRES and Post-SRES scenarios, are generally able to accountfor these underlying anti-correlation effects. Thus, the following analysis assumesthat an analysis of the SRES and Post-SRES scenarios can provide insights aboutreal world dynamics in regard to whether inherent process based anti-correlationsof emissions are so dominant, that the unity rank correlation assumption at givenaggregation levels would be invalidated.

The question is, therefore, whether any negative rank correlations are apparentat the aggregation level considered here, namely the 4 SRES world regions. Forthe pool of existing SRES and Post-SRES scenarios that are used, no negative rankcorrelations between fossil fuel CO2 and any other gases’ emissions are apparentat this stage of aggregation by sources and regions (see Figure 7 and Appendix B).The rank correlation between fossil fuel CO2 and ‘Other CO2’ or ‘N2O total’ isbasically zero or rather small, while rank correlations with other gases are positive,especially for the ASIA and ALM region.

Between fossil fuel CO2 and the land-use and agriculture dominated ‘OtherCO2’ and ‘N2O’ emissions, there is little or no rank correlation. In other words, inthe underlying SRES and post SRES data set, the sources of these emissions arelargely unrelated. The primary reason for this is that ‘Other CO2’ sources are atpresent dominated by tropical deforestation (Fearnside, 2000). Another reason whyexisting scenarios with low fossil fuel CO2 emissions do not necessarily correspondto large reductions in deforestation emissions or large net sequestration appears tobe that some modeling groups assume different policy mixes or different root causesof deforestation–potentially out of reach for climate policies.

In summary, the validity of the EQW approach is not limited as long as it isonly applied at aggregation levels, where negative rank correlations are generallynot evident, as is the case in this study. The fact that there are inherent, process


Figure 7. Rank correlations within the pool of existing SRES/Post-SRES scenarios between fossilfuel CO2 emissions and other greenhouse gas emissions and aerosols (columns) for the 4 SRES Worldregions (rows). The Kendall rank correlation (solid line), its mean from 2010 to 2100 (μ) and theSpearman rank correlation (dotted lines) are given (see Appendix B).

based anti-correlations of certain emissions at local or more subsource-specificlevel(s), does not invalidate this unity rank correlation assumption, as long as theseunderlying anti-correlations are not dominant.

The differing population assumptions of the underlying scenarios might appearto be, at first sight, a reason for the positively rank correlated emissions acrossdifferent gases. A scenario that assumes high population growth is likely to predicthigh human-induced emissions across all gases. However, a closer look at per-capita(instead of absolute) emissions shows that differing population assumptions are notthe reason for the positively rank correlated emission levels nor the large variationof absolute emissions. Rank correlations across the different gases on a per-capitabasis (a) are generally non-negative and (b) are not uniformly lower or higheracross all regions and gases than do rank correlations that are based on absoluteemissions. On average, these per capita rank correlations are only marginally lowerthan rank correlations based on absolute emissions. Specifically, the change of themean Kendall rank correlation index over 2010 to 2100 is insignificantly differentfrom zero (−0.008) when averaged over all gases. Maximal changes are +0.07and −0.11 for some gases (standard deviation of 0.043), if per-capita emissions areanalyzed instead of absolute emissions (cf. Figure 7).


Given the absence of negatively rank correlated emissions, the seeming dis-advantage of the EQW approach, namely that it assumes unity rank correlationbetween fossil CO2 emissions and those of other gases, might actually be an ad-vantage. Since the EQW approach is primarily designed to create new familiesof intervention pathways, correlating reduction efforts between otherwise uncor-related greenhouse gas sources might be a sensible characteristic. In other words,for those sources that are not correlated with fossil fuel CO2 emissions, namelyland-use dominated and agricultural emissions, the EQW approach suggests thata climate-policy-mix might tackle these sources in parallel to tackling fossil fuelemissions. Given that some policy options are available to reduce emissions in theland-use sector (see e.g. Pretty et al., 2002; see e.g. Carvalho et al., 2004)8 it wouldseem very likely that the more a reduction effort is put into reducing fossil fuelrelated emissions, the more a parallel reduction effort will be put into reducingland-use related emissions as well.

5.1.2. Assuming a Certain Probability of Underlying Scenarios?The application of some statistical tools within the EQW method assumes equalvalidity of each of the 54 scenarios within the underlying pool. This assumption,however, does not affect the outcome. As the following results show, the EQWmethod is rather robust to the relative ‘probability’ (weighting) within the scenariopool. Thus, the EQW method is largely independent of the assumed likelihood ofsingle scenarios.

The sensitivity of the EQW method to different weightings of the underlyingscenarios has been analyzed as follows. Four sensitivity runs have been performed.In each of them, members of one of the four IPCC scenarios families A1, A2,B1 and B2 have been multiplied three times. In effect, the original 54 plus themultiplied scenarios were then analyzed to derive the ‘distributions of possibleemission levels’, as outlined above (3.1). Keeping other parts of the EQW method thesame, intervention pathways were derived for global-mean temperature peaking at2 ◦C above the pre-industrial level. The results show that the pathways’ sensitivitiesto the weighting are rather small. Obviously, if a scenario’s frequency or weight-factor is changed, slightly different emissions pathways will result, since basicallyall scenarios differ with respect to relative gas and regional shares (see Table IV).

Obviously, assuming a different set of scenarios altogether in order to derive thedistribution of possible emission levels might change the outcome considerably.

It should be kept in mind that the EQW method is not designed to determine howlikely it might be that future emissions will be below a certain level. Similar to themedians calculated by Nakicenovic et al. (1998) for the IPCC database, the derived‘distributions of possible emission levels’ are by no means probability estimations(cf. e.g. Grubler and Nakicenovic, 2001). If, however, one would have a set ofscenarios with a well defined likelihood for each of them, then more far reachingconclusions could be drawn instead of designing normative scenarios, as is donehere.


TABLE IVSensitivity analysis with respect to the underlying SRES scenario family frequencies. The commonclimate target ‘peaking below 2◦C’ is prescribed for and met by all 5 pathways assuming a climatesensitivity of 2.8◦C (7 AOGCM ensemble mean). Whereas the first pathway (EQW-P2T) was derivedby using the underlying data pool of 54 unique scenarios, the four sensitivity pathways were derivedby multiplying the frequency of A1, A2, B1 or B2 scenario family members three times (3 × A1 to3 × B2). Shown are the emission levels in 2050 compared to 1990 levels for different gases (a) andregions (b) and the annual reduction rate for OECD fossil CO2 emissions (c)

EQW-P2T 3 × A1 3 × A2 3 × B1 3 × B2(%) (%) (%) (%) (%)

(a) Gas-by-gas results for region “World”

(Emission levels in 2050 compared to 1990)

Fossil CO2 73 68 71 78 82

CH4 91 93 87 96 82

N2O 74 78 75 73 74

F-gases 67 64 58 71 64

6-gas 76 74 75 80 80

6-gas (incl. ‘Other CO2’) 61 60 60 65 65

(b) Regional results for “6-gas” (incl. ‘Other CO2’)

(Emission levels in 2050 compared to 1990)

OECD 37 34 35 41 43

REF 11 13 8 18 5

ASIA 110 110 112 109 118

ALM 85 78 80 90 85

World 61 60 60 65 65

(c) Driver path

(Annual reduction rate)

OECD fossil CO2 −3.3 −3.6 −3.6 −2.9 −2.6

5.1.3. Sensitivity to Lower Range ScenariosIf the EQW method produces a new emissions pathway near to or slightly outside therange of existing scenarios, there is a high sensitivity to scenarios in the underlyingdata base that are at the edge of the existing distribution. Certain measures canand are applied to limit this sensitivity, and its undesired effects, by (a) using anappropriate kernel-width to derive the ‘distribution of possible emission levels’ (seeSection 3.1), (b) enlarging the pool of underlying scenarios by explicit interventionscenarios at the lower edge of the distribution, namely by the inclusion of Post-SRESstabilization scenarios, while at the same time (c) restricting the pool to scenariosof widely accepted modeling groups with integrated and detailed models.

Clearly, entering ‘unexplored’ terrain with this approach is only a second bestoption in the absence of fully developed scenarios for the more stringent climatetargets. Ideally, the EQW method would be applied on a large pool of scenarios


including those with the most stringent climate targets. Such fully developed mit-igation scenarios might be increasingly available in the future. For example, newMESSAGE and IMAGE model runs (Nakicenovic and Riahi, 2003; van Vuurenet al., 2003) and forthcoming multi-gas scenarios developed within the EnergyModeling Forum EMF-21 (see e.g. de la Chesnaye, 2003) could build the basis ofupdated EQW pathways.

5.1.4. Regional Emissions & Future Commitment AllocationsGeo-political realities, the historic responsibility of different regions, their ability topay, capability to reduce emissions, vulnerability to impacts as well as other fairnessand equity criteria will inform the global framework for the future differentiation ofreduction commitments. Thus, splitting up a global emissions pathway and choosinga commitment differentiation is not solely a scientific or economic issue, but rathera (sensitive) political one.

Regionally different emission paths result from the application of the EQWmethod to the 4 SRES regions. This is a direct consequence of the regional emissionshares within the pool of underlying SRES / Post-SRES scenarios as well as possiblyregionally differentiated departure years from the median (see Section 3.2). Thus,the EQW method is not, in itself, an emission allocation approach based on explicitdifferentiation criteria. The method captures the spectrum of allocations in the poolof underlying existing scenarios and allows for some flexibility by setting regionallydifferentiated departure years for example.

Under default assumptions, the derived emissions pathways entail an increasingshare of non-Annex I emissions independent of the climate target (Figure 8). Thisis in accordance with many of the approaches for the differentiation of futurecommitments (den Elzen, 2002; Hohne et al., 2003). Nevertheless, a sensitivityanalysis with different climate parameters, departure years and possibly differentquantile paths for different regions allows making important contributions in thediscussion on future commitments. In addition, EQW pathways can be used as inputfor detailed emission allocation analysis tools, such as FAIR (den Elzen and Lucas,2005), in order to obtain assessments of future climate regime proposals that areconsistent with certain climate targets.

5.1.5. Baseline Independency & Absence of Socio-Economic PathsIn line with the most popular previous mitigation pathways, the derived pathwaysdo not attempt to reflect a certain socio-economic development pathway. The socio-economic characteristics of a future world can hardly be derived by walking alongcertain quantiles of the distributions of GDP development, productivity, fertility,etc. As pointed out by Grubler and Nakicenovic (2001): “Socioeconomic variablesand their alternative future development paths cannot be combined at will and arenot freely interchangeable because of their interdependencies. One should not, forexample, create a scenario combining low fertility with high infant mortality, or zeroeconomic growth with rapid technological change and productivity growth – since


Figure 8. The regional implications of EQW emissions pathways for a peaking at 470 ppm CO2eq (a,c) and 555 ppm CO2eq (b, d). Whereas Annex I countries (bright slices OECD and REF) caused thelion’s share of emissions in the past, the more populated non-Annex I regions (darker slices ALM andASIA) are projected to cause higher emissions in the future under the derived intervention pathways.This characteristic holds for fossil CO2 emissions (top row) and the aggregated set of greenhouse gasemissions including land-use related CO2 emissions (lower row).

these do not tend to go together in real life any more than they do in demographicor economic theory.”

The lack of a socio-economic description of the future world is a disadvantageof the EQW method in comparison to intervention scenarios derived according tofully developed scenario approaches with or without cost-optimization (see meth-ods three and four as described in Section 1). However, the baseline independencyand more general nature of the presented EQW pathways allows for a more ubiq-uitous application and for further comparative analyses in regard to the emissionimplications of certain climate targets. Alternatively, a restriction of the underlyingpool of scenarios to one specific scenario family would allow the derivation ofbaseline-dependent intervention pathways.

5.1.6. Land-Use-Change Related Emissions – A Word of CautionThe following paragraph is a general word of caution on the interpretation ofland-use related sinks and emissions within the EQW pathways. There are severaldistinctive characteristics of land-use versus energy related emission reductions


that complicate their appropriate reflection and interpretation in intervention sce-narios. Firstly, in regard to land-use related CO2 net removals (cf. Figure 6, leftcolumn, graph b): sequestration might not bind the carbon for a very long time.Today’s biospheric sinks might turn into tomorrow’s sources. Therefore, enhance-ment of (temporary) biospheric CO2 sequestration is not equivalent to restrictingfossil fuel related emissions under a long-term perspective (Lashof and Hare, 1999;Kirschbaum, 2003; Harvey, 2004). Secondly, the root causes of land-use relatedemissions are even more complex for land-use emissions than for energy relatedemissions (Carvalho et al., 2004). Thus, without a carefully balanced policy mix,negative side effects for biodiversity, watershed management, and local commu-nities might offset carbon uptake related benefits under a broader sustainabilityagenda. Thirdly, land-use related emission allowances under the current rules ofthe Kyoto Protocol are largely windfall credits that do not reflect additional seques-tration or real emission reductions. Fourthly, ‘natural’ variability of the biosphericcarbon stock poses risks for the regime stability of an emission control architecture.Given these issues, the presented results should be regarded with care. In particular,they should not be misinterpreted as a call for the advancement of sink related emis-sion allowances in the way followed so far under the international climate changeregime.

5.1.7. Studying Alternative Gas-to-Gas and Timing StrategiesSome studies analyze the relative merits of focusing reduction efforts on somespecific radiative forcing agents, such as methane and ozone precursors (see e.g.Hansen et al., 2000). Deriving alternative emissions pathways that reflect differinggas-to-gas mitigation strategies for the same climate target might thus be a desirablepart of a broader sensitivity analysis. The method could be extended by applyingdifferent ‘quantile paths’ to different gases, not only different regions. Such a ‘Dif-ferentiated Quantile Walk’ method could allow systematically analyzing differentmitigation strategies. For example, methane and nitrous oxide emissions could bereduced according to a ‘quantile path’ that is equivalent to a 3% annual reduction offossil fuel CO2 emissions, while in fact fossil CO2 is reduced by only 2% annually(cf. Section 3.2).

The flexible nature of the EQW method allows deriving pathways with dif-ferent timings for emission reductions. As already demonstrated by the presen-tation of stabilization and peaking profiles, emissions pathways for various tar-get paths can be derived. Depending on the definition of the index or quantilepaths (Sections 3.2 and 3.3), emissions pathways can be designed that result in amonotonic increase of temperature or CO2 concentrations up to a final target levelwith stabilization thereafter or subsequent dropping (e.g. overshooting (see e.g.Wigley, 2003b) or peaking profiles). Furthermore, the possibility to freely definethe departure year for various regions allows future studies to undertake sensitiv-ity studies contributing to the debate on ‘early action’ versus ‘delayed response’(cf. Section 1).


5.1.8. Probabilistic FrameworkThe EQW method can be used to systematically explore the effect of uncertaintiesin the climate system upon emission implications in a probabilistic framework (seeSection 4.2). A probabilistic framework is important to allow for the definition ofan optimal hedging strategy against dangerous climate change. Any ‘best guess’parameter model runs might lead to a systematic underestimation of optimal re-duction efforts. A ‘best guess’ answer in regard to the emission implications willonly imply a 50% certainty to actually achieve the climate target. Under both a‘cost-benefit’ and a ‘normative target’ policy framework, policymakers might wantto design more ambitious reduction policies in order to hedge against the pos-sibility of overshooting the target or against the possibility of costly mid-courseadjustments. Specifically, fossil fuel related CO2 emissions (allowances) in OECDcountries would have to decrease by 3% annually after 2010, with emissions fromother sources and regions corresponding to the same quantile path, in order to limitthe probability of overshooting 2 ◦C to 25% to 77% (see Table III). 3% annualemission reductions may not be sufficient, if one wishes to ensure that the warmingtrajectory never exceeds the 2 ◦C target with a higher certainty.

5.2. DISCUSSION OF LIMITATIONS ARISING FROM THE UNDERLYING DATABASE

The derived emissions pathways will inevitably share some of the limitations ofthe underlying pool of existing scenarios. In the following, quantitative and qual-itative limitations of the scenario database are briefly highlighted (Section 5.2.1).Subsequently, one of the qualitative limitations, namely the potentially inadequatereflection of land-use related non-CO2 emissions, is discussed in more detail anda comparison to recently developed cost-optimized mitigation scenarios is drawn(Section 5.2.2).

5.2.1. Quantitatively and Qualitatively Limited Pool of ScenariosThe 54 SRES and Post-SRES scenarios used in this study provide a solid basis forthe derived emissions pathways. However, as the number and quality of long-termemission scenarios will increase in the future, thanks to ongoing concerted researchefforts, the quality of and level of detail in the derived EQW pathways should alsobe enhanced. Most importantly, the sensitivity to single scenarios would be loweredby basing the EQW method on more scenarios, provided that these scenarios arein turn based on sound and independently researched studies of mitigation poten-tials. Lowering this sensitivity to single scenarios seems especially warranted forthe lower emissions pathways (cf. Section 5.1.3). Going beyond the mere numberof scenarios, an extended time horizon, and higher detail in terms of (standard-ized) regional and source-specific information in the scenarios, would enhance theusefulness of derived EQW pathways.

Furthermore, some qualitative limitations within the set of used SRES and Post-SRES should be kept in mind when using the presented EQW pathways. For


example, the SRES and Post-SRES scenarios were developed prior to the year2000. Thus, the original scenarios and the derived intervention emissions pathwaysmight not fully match actual emissions up to the present day, although differencesseem to be limited (van Vuuren and O’Neill, submitted).

5.2.2. A Comparison with Recently Derived Multi-Gas ScenariosThe Post-SRES scenarios within the underlying pool might have one shortcomingin common: all those scenarios were primarily focused on energy related reductionpotentials with little details on other sectors and sources, such as land-use relatednon-CO2 emissions (see e.g. Jiang et al., 2000; Morita et al., 2000).

To explore this potential limitation, a comparison with some of the recent mit-igation scenarios has been done, which have been developed in relation to a co-ordinated modeling effort in the context of the Energy Modeling Forum (de laChesnaye, 2003). These scenarios are designed to find cost-optimized multi-gasreduction paths with a more sophisticated representation of non-CO2 greenhousegases than captured by most previous scenarios. For that purpose, a standardizeddatabase of mitigation measures for the most important sources of CH4, N2O andhalocarbons and halogenated compounds was developed. The various modelinggroups used different approaches, ranging from macro-economic models to moretechnology-rich and integrated assessment ones. For the most important sources ofCH4 and N2O, i.e., agricultural and land use-related sources, the measures capturedin the range of 10–50% of total emissions at cost levels of 200 US$/tC. For energyand industrial sources, the potential reductions were higher-and ranged up to nearly100%. After incorporating the non-CO2 reduction options into the models, cost-optimal reduction scenarios for a radiative forcing stabilization at 4.5 W/m2 werederived. Some modeling teams, such as the IMAGE group and the developers ofMERGE, also developed scenarios for other climate targets involving in some casesthe full range of land use and agriculture emissions (see e.g. Manne and Richels,2001; van Vuuren et al., 2003).

In the following, EMF-21 multi-gas scenarios of the participating modelinggroups9 are compared to an EQW emissions pathway (see Figure 9). All pathwaysand scenarios are designed to achieve a moderately ambitious climate target, namelyto lead to a maximal radiative forcing of 4.5W/m2. In general, the EQW pathwayfalls well within the range spanned by the EMF-21 scenarios. For CO2 and N2O,the EQW result is in fact close to the EMF-21 median. For CH4, the EMF-21median seems to be lower than the EQW result indicating that specific attentionto reduction possibilities of CH4 can result in lower CH4 emissions. Differencesbetween emission trajectories of EMF-21 and the EQW pathway are even reduced,if the set of emission sources were standardized. In particular for N2O and to somedegree for CH4, the EMF-21 results are rather scattered already in the historic year2000 as some models have not included all emission sources. In addition, differentdefinitions are used for land-use related N2O emissions in terms of what constitutesthe anthropogenic part.


a. fossil CO2

b. CH4

c. N2O

0

20

40

60

80

100

120

140

0

5

10

15

20

2000 2020 2040 2060 2080 2100 2120 21400

1

2

3

4

5

6

7

8

9

CtG( s

nois si

mEO

2 )

CtG( s

nois si

mEO

2 )e

CtG( s

nois si

mEO

2 )e

SRES & post-SRES poolEMF-21EQW P-4.5W/m2

Figure 9. Fossil carbon dioxide (a) and methane (b) and nitrous oxide (c) emissions of an EQWpathway (solid black line), the IPCC SRES and Post-SRES scenarios used as underlying scenariopool in this study (solid grey lines) and recently developed multi-gas scenarios under the EMF-21(dashed black lines). The EQW pathway and the EMF-21 scenarios are designed to lead to a maximalradiative forcing of 4.5 W/m2. Discussion see text.


The main conclusion is that the presented EQW pathways seem to be alreadysimilar to those found in more detailed modeling studies that account for specificmitigation options as suggested by EMF-21 work. At this rather moderate climatetarget of 4.5W/m2, the different emissions pathways do not widely diverge. For allgases, emissions end up in 2100 slightly below current emission levels. This is boththe case in the EQW and the EMF-21 results.

It would be an improvement, though, to extend the sample of scenarios that EQWdraws from by including these EMF-21 scenarios and other elaborated multi-gasscenarios in the underlying scenario pool, as they become available for a stan-dardized set of emission sources. Thereby the EQW method could capture a widerrange of non-CO2 mitigations options. The ‘distribution of possible emission levels’within EQW will become less dependent on differences in driving forces andmodels (that are currently likely to dominate the range) and more dependent onthe potential for emission reductions among the different gases and their relativecosts.

6. Conclusions and Further Work

This study proposes a method to derive emissions pathways with a consistent treat-ment of all major greenhouse gases and other radiative forcing agents. For example,multi-gas emissions pathways can be derived for various climate target indicatorsand levels, such as stabilization of CO2 concentrations at 450 ppm or for peaking ofradiative forcing at 2.6 W/m2 (≈470 ppm CO2 equivalence) above the pre-industriallevel. The proposed EQW method has various advantages, such as being flexibleand applicable to various research questions related to Article 2 of the UNFCCC.For example, derived EQW emissions pathways can be used to perform transientclimate impact studies as well as to study emission control implications associatedwith certain climate targets. Of course, the EQW method can only fill a niche, andcannot replace other more mechanistic multi-gas approaches, e.g. cost optimiza-tion procedures. On the contrary, the EQW method is crucially dependent on andbuilds on a large pool of existing and fully developed scenarios. Thus, the derivedregion-specific and gas-specific emission paths respect the ‘distributions of possibleemission levels’ as they were outlined before by many different modelling groups.Another characteristic of the EQW pathways is that they are, to a large extent,baseline independent. Thus, the EQW pathways could be attractive for designingcomparable climate impact and policy implication analyses.

Achieving climate targets that account for, say, the risk of disintegrating icesheets (Oppenheimer, 1998; Hansen, 2003; Oppenheimer and Alley, 2004) or forlarge scale extinction risks (Thomas et al., 2004) almost certainly requires sub-stantial and near term emission reductions. For example, to constrain global-meantemperatures to peaking at 2 ◦C above the pre-industrial level with reasonable cer-tainty (say >75%) would require emission reductions of the order of 60% below


1990 levels by 2050 for the GWP-weighted sum of all greenhouse gases (cf. peakingpathway I in Table III). If the start of significant emission reductions were furtherdelayed, the necessary rates of emissions reduction rates were even higher, if therisk of overshooting certain temperature levels shouldn’t be increased (den Elzenand Meinshausen, 2005; Meinshausen, 2005). Thus, since more rapid reductionsmay require the premature retirement of existing capital stocks, the cost of anyfurther delay would be increased, probably non-linearly. There are a number ofother reasons, why one might want to avoid further delay. Firstly, future genera-tions face more stringent emission reductions while already facing increased costsof climate impacts. Secondly, the potential benefits of ‘learning by doing’ (Arrow,1962; Gritsevskyi and Nakicenovi, 2000; Grubb and Ulph, 2002) were limited dueto the more sudden deployment of new technology and infrastructure. Thirdly, afurther delay of mitigation efforts risks the potential foreclosure of reaching certainclimate targets. Thus, a delay might be particularly costly if, for example, the cli-mate sensitivity turns out to be towards the higher end of the currently assumedranges (cf. Andronova and Schlesinger, 2001; Forest et al., 2002; Knutti et al.,2003).

So far, the development of optimal hedging strategies against dangerous climatechange has been hampered by the absence of a method to generate flexible andconsistent multi-gas emissions pathways. In this regard, the EQW method could bean important contributor towards the development of more elaborate and compre-hensive climate impact and emission control studies and policies in a probabilisticframework.

Acknowledgements

First of all, the authors are thankful to the most helpful review comments by Fran-cisco de la Chesnaye and an anonymous reviewer. The authors are indebted to themodeling groups participating in EMF-21, whose data, compiled by G.J. Blanfordhas been used for comparison. The authors would also like to thank Claire Stock-well and Vera Tekken for magnificent editing support and Nicolai Meinshausen formost valuable help on statistics. Furthermore, we truly benefited from inspiring dis-cussions with Dieter Imboden, Marcel Berk, Michiel Schaeffer, Bas Eickhout andAdrian Muller. Malte Meinshausen would also like to thank the whole RIVM teamfor their hospitality during a four-month research visit. As usual, shortcomings inthis study remain the responsibility of the authors.

Appendix A

This Appendix A entails a description of (a) the employed simple MAGICC and(b) the assumptions made in regard to solar and volcanic forcings.


A.1. THE MODEL

(a) For the computation of global mean climate indicators, the simple climate modelMAGICC 4.1 has been used.10 MAGICC is the primary simple climate model thathas been used by the IPCC to produce projections of future sea level rise and global-mean temperatures. The description in the following paragraph is largely based onWigley (2003a). Information on earlier versions of MAGICC has been publishedin Wigley and Raper (1992) and Raper et al. (1996). The carbon cycle model is themodel of Wigley (1993), with further details given in Wigley (2000) and Wigley andRaper (2001). Modifications to MAGICC made for its use in the IPCC TAR (IPCC,2001) are described in Wigley and Raper (2001, 2002) and Wigley et al. (2002).Additional details are given in the IPCC TAR climate projections chapter 9 (Cubaschet al., 2001). Sea level rise components other than thermal expansion are describedin the IPCC TAR sea level chapter 11 (Church et al., 2001) with an exception inrelation to the contribution of glaciers and small ice caps as described in Wigley(2003a). Gas cycle models other than the carbon cycle model are described in theIPCC TAR atmospheric chemistry chapter 4 (Ehhalt et al., 2001) and in Wigley etal. (2002). The representation of temperature related carbon cycle feedbacks hasbeen slightly improved in comparison to the MAGICC version used in the IPCCTAR, so that the magnitude of MAGICC’s climate feedbacks are comparable tothe carbon cycle feedbacks of the Bern-CC and the ISAM model (see Box 3.7 inPrentice et al., 2001).11

A.2. PARAMETER CHOICES

Ensemble mean outputs of this simple climate model are the basis for all presentedcalculations in this study. An exception are the probabilistic results of Section 4.2,where MAGICC TAR default parameters were complemented by the probabilitydensity distributions of different authors’ estimates of climate sensitivity, to obtainprobabilistic forecasts (see e.g. Andronova and Schlesinger, 2001; Forest et al.,2002; Gregory et al., 2002; Knutti et al., 2003; Murphy et al., 2004). The ensembleoutputs are computed as means of seven model runs. In each run, 13 model parame-ters of MAGICC are adjusted to optimal tuning values for seven atmospheric-oceanglobal circulation models (AOGCMs). This ‘ensemble mean’ procedure is widelyused in the IPCC Third Assessment Report and described in Appendix 9.1 (seeTable 9.A1 in Cubasch et al., 2001; Raper et al., 2001). By using this ‘ensem-ble mean’ procedure, the implicit assumptions in regard to climate sensitivity andocean diffusivity are based on the seven AOGCMs. The mean climate sensitiv-ity for those 7 AOGCMs models is 2.8 ◦C per doubled CO2 concentration levels(median is 2.6 ◦C). Clearly, if emission scenarios are derived with single modeltunings or different climate sensitivities then different emission paths will be foundto correspond to any given climate target, reflecting the underlying uncertainty in


the science. In general, the CO2 concentration and radiative forcing scenarios areless model parameter dependent than the temperature focused scenarios.

A.3. CAVEATS

MAGICC is probably the most rigorously tested model among the simple climatemodels. Nevertheless, general caveats apply as well. There are still uncertaintiesin regard to many aspects of our understanding of the climate system, appropriatemodel representations and parameter choices, such as for gas cycles and their in-teractions, temperature feedbacks on the carbon cycle, ocean mixing, the climate’ssensitivity etc. For example, large uncertainties persist in regard to the radiativeforcing due to reactive gas emissions, such as NOx. In this case, MAGICC usessimple algorithms developed for the IPCC Third Assessment Report (see Wigleyet al., 2002 for further information on this). However, in most cases, the effect ofthese uncertainties on long-term global-mean temperature projections is relativelysmall. The large uncertainties in regard to indirect aerosol forcing are another ex-ample. Obviously, a best estimate parameter as used in the IPCC Third AssessmentReport calculations and in this study does not reflect these uncertainty ranges. How-ever, at the global mean level the effect of aerosol forcing uncertainties is limited forlong-term projections as aerosol precursor emissions are expected to decline overthe 21st century, as discussed in (Wigley and Raper, 2002). The major source ofuncertainty for long-term global-mean temperature projections, the climate sensi-tivity, has been explored in this study (see Section 4.2, and A.2). Future applicationswill benefit from a truly probabilistic framework (cf. Section 5.1.8).

A.4. NATURAL FORCINGS

Historic solar and volcanic forcings have been assumed, as presented in the IPCCTAR and according to Lean et al. j (1995) and Sato et al. (1993), respectively (seeFigure 6–8 in Ramaswamy et al., 2001). Recent studies suggested that an up-scalingof solar forcing might lead to a better agreement of historic temperature records (e.g.Hill et al., 2001; North and Wu, 2001; Stott et al., 2003). In accordance with the bestfit results by Stott et al. (2003, Table II), a solar forcing scaling factor of 2.64 hasbeen assumed for this study. Accordingly, volcanic forcings from Sato et al. (1993)have been scaled down by a factor 0.39 (Stott et al., 2003, Table 2). However, thereis considerable uncertainty in this regard and it should be noted that mechanismsfor the amplification of solar forcing are not yet established (Ramaswamy et al.,2001, section 6.11.2; Stott et al., 2003). Future solar and volcanic forcings havebeen assumed in accordance with the mean forcings over the past 22 and 100 yearsrespectively, i.e. +0.16 W/m2 for solar and −0.35 W/m2 for volcanic forcing andscaled as described above.12


Appendix B

Spearman rank correlations ‘SRCorr’ between fossil fuel CO2 emissions and theemissions of gas g at time t are given as:

SRCorrg,t = (RfCO2,t − μ)(Rg,t − μ)

σ 2

where Rg,t is the vector of rank indexes for each scenario at time t for gas g, RfCO2,tis the vector of rank indexes for each scenario at time t for fossil CO2 emissions,μ is the mean of all ranks (in this case half the number of scenarios +0.5) and σ

is the standard deviation of the rank indexes. Another indicator is the Kendall rankcorrelation indicator given as:

KRCorrg,t

= 1

n(n − 1)

n∑i=1

(n∑

s=1

sign(efCO2,s − efCO2,i )sign(eg,s − eg,i )

)with s �= i

where n is the number of scenarios, eg,s the emission of gas g for scenario s andwhere the function ‘sign(..)’ returns −1 for negative and +1 for positive differencesin emissions between two scenarios.

Notes

1The United Nations Framework Convention on Climate Change (UNFCCC) is available onlineat http://unfccc.int/resource/docs/convkp/conveng.pdf. Its status of ratification can be accessed athttp://unfccc.int/resource/conv/ratlist.pdf.

2The 40 IPCC SRES scenarios were used as presented in the IPCC SRES database (version 1.1),available at http://sres.ciesin.org/final data.html, accessed in March 2004.

3For details on the six modelling groups (AIM, ASF, IMAGE, MARIA, MESSAGE, MiniCAM)that quantified the 40 SRES and 14 Post-SRES scenarios used, see Box TS-2 and Appendix IVin Nakicenovic and Swart (2000), available online at http://www.grida.no/climate/ipcc/emission/,accessed in May 2004.

4However, even among the recently developed EMF-21 scenarios, only very few suggest that N2Oemissions might fall much below current levels (cf. Figure 9) as most of the spread among EMF-21scenarios seems to stem from different N2O source inclusions and definitions, not from reductionpotentials.

5This does not mean that overall terrestrial carbon stocks are restored to pre-industrial levels.Elevated CO2 concentrations are thought to increase the total amount of terrestrial biotic carbonstocks. Thus, despite a partially counterbalancing effect due to climate change (Cramer et al., 2001),terrestrial carbon stocks are likely to increase above levels in 1850, if the directly human-inducedcarbon uptake due to future afforestation and reforestation programmes is equivalent to the directlyhuman-induced deforestation related emissions since 1850.

6Annex I refers to the countries inscribed in Annex I of the United Nations Framework Conventionon Climate Change and corresponds to the IPCC SRES regions OECD and REF. Consequently, non-Annex I corresponds to the IPCC SRES regions ASIA and ALM.

7In the future, the negative radiative forcing from sulphur aerosols is likely to become much lessimportant according to the majority of SRES and post-SRES scenarios, which expect reduced sulphuremissions as a consequence of air pollution control policies.


8Given that fossil CO2 emissions have been used as the ‘driver path’, correlations have beenanalyzed between fossil CO2 emissions and other radiative forcing agent emissions. However, cor-relations among different sets of gases can be more complex, particularly when analyzed on a lessaggregated level. For example, Wassmann et al. (2004) showed that in the rice-wheat system in Asiathere are clear antagonisms between measures that reduce methane and nitrous oxide: reducing oneoften leads to increases in the other.

9The participating modelling groups for EMF-21 are AIM, AMIGA, COMBAT, EDGE, EPPA,FUND, GEMINI-E3, GRAPE, GTEM, IMAGE, IPAC, MERGE, MESSAGE, MiniCAM, SGM,WIAGEM. The work of these groups is gratefully acknowledged. Emission scenarios of these mod-elling groups are plotted in Figure 9.

10MAGICC 4.1 has been developed by T.G.L. Wigley, S. Raper and M. Hulme and is available athttp://www.cgd.ucar.edu/cas/wigley/magicc/index.html, accessed in May 2004.

11This improvement of MAGICC only affects the no-feedback results. When climate feedbackson the carbon cycle are included, the differences from the IPCC TAR are negligible.

12The alternative, to leave natural forcings out in the future, is not really viable, since the model hasbeen spun up with estimates of the historic solar and volcanic forcings. Assuming the solar forcing tobe a non-stationary process with a cyclical component and assuming that the sum of volcanic forcingevents can be represented as a Compound Poisson process, it seems more realistic to apply the recentand long-term means of solar and volcanic forcings, respectively, for the future.

13The four SRES World regions are: OECD – Members of the OECD in 1990; REF – Countriesundergoing economic reform, namely Former Soviet Union and Eastern Europe; ASIA – Asia; ALM –Africa and Latin America. See Appendix III in Nakicenovic and Swart (2000) for a country-by-countrydefinition of the groups.

14Since the introduction of the GWP concept (1990), it has been the subject of continuous scientificdebate on the question of whether it provides an adequate measure for combining the different effectson the climate system of the different greenhouse gases (Smith and Wigley, 2000a; Smith and Wigley,2000b; Manne and Richels, 2001; Fuglestvedt et al., 2003). The GWP concept is very sensitive to thetime horizon selected, and can only partially take into account the impacts of the different lifetimes ofthe various gases. Economists currently criticise GWP for not taking economic efficiency into account.However, despite its limitations, the GWP concept is convenient and has been widely used in policydocuments such as the Kyoto Protocol. To date, no alternative measure has attained a comparablestatus in policy documents.

15Data on the ‘S’ profiles is available at http://cdiac.ornl.gov/ftp/db1009/, accessed in March 2004.16Note that the 14 Post-SRES scenarios used in this study have been selected from those mod-

elling groups that provided the 40 SRES scenarios as well, namely AIM, MESSAGE, IMAGE, ASF,MiniCAM, and MARIA (see as well endnote 3).

17As shown later, the EQW methodology allows one to easily deriving profiles for different targetvariables, such as CO2 concentrations, global mean temperatures, radiative forcing or sea level, andfor different profile shapes, such as stabilization, overshooting or peaking scenarios.

18The peak concentration is shown for the 7 AOGCM ensemble mean. Due to the temperaturefeedback on the carbon cycle, the actual peak concentration varies slightly depending on the assumedclimate sensitivity.

References

Anderson, T. L., Charlson, R. J., Schwartz, S. E., Knutti, R., Boucher, O., Rodhe, H., and Heintzenberg,J.: 2003, ‘Climate forcing by aerosols a hazy picture’, Science 300, 1103–1104.

Andronova, N. G. and Schlesinger, M. E.: 2001, ‘Objective estimation of the probability densityfunction for climate sensitivity’, Journal of Geophysical Research-Atmospheres 106, 22605–22611.


Arnell, N. W., Cannell, M. G. R., Hulme, M., Kovats, R. S., Mitchell, J., Nicholls, R. J., Parry, M. L.,Livermore, M. T. J., and White, A.: 2002, ‘The consequences of CO2 stabilistation for the impactsof climate change’, Climatic Change 53, 413–446.

Arrow, K.: 1962, ‘The Economic Implications of Learning-by-Doing’, Review of Economic Studies29, 155–73.

Azar, C.: 1998, ‘The timing of CO2 emissions reductions: the debate revisited’, International Journalof Environment and Pollution 10, 508–521.

Azar, C. and Rodhe, H.: 1997, ‘Targets for stabilization of atmospheric CO2’, Science 276, 1818–1819.Bowman, A.W. and Azzalini, A.: 1997, Applied Smoothing Techniques for Data Analysis, Oxford

University Press, Oxford, UK.Cai, W. J., Whetton, P. H., and Karoly, D. J.: 2003, ‘The response of the Antarctic Oscillation to

increasing and stabilized atmospheric CO2’, Journal of Climate 16, 1525–1538.Carvalho, G., Moutinho, P., Nepstad, D., Mattos, L., and Santilli, M.: 2004, ‘An amazon perspective on

the forest-climate connection: opportunity for climate mitigation, conservation and development?’Environment, Development and Sustainability 6, 163–174.

Church, J. A., Gregory, J. M., Huybrechts, P., Kuhn, M., Lambeck, K., Nhuan, M. T., Qin, D., andWoodworth, P. L.: 2001, ‘Changes in sea level’, in Houghton, J. T., Ding, Y., Griggs, D. J., Noguer,M., van der Linden, P. J., Dai, X., Maskell, K., and Johnson, C.A. (eds.), Climate Change 2001:The Scientific Basis, Cambridge University Press, Cambridge, UK, pp. 892.

Cramer, W., Bondeau, A., Woodward, F. I., Prentice, I. C., Betts, R. A., Brovkin, V., Cox, P. M., Fisher,V., Foley, J. A., Friend, A. D., Kucharik, C., Lomas, M. R., Ramankutty, N., Sitch, S., Smith, B.,White, A., and Young-Molling, C.: 2001, ‘Global response of terrestrial ecosystem structure andfunction to CO2 and climate change: results from six dynamic global vegetation models’, GlobalChange Biology 7, 357–373.

Cubasch, U., Meehl, G. A., Boer, G. J., Stouffer, R. J., Dix, M., Noda, A., Senior, C. A., Raper, S.,and Yap, K. S.: 2001, ‘Projections of future climate change’, in Houghton, J.T., Ding, Y., Griggs,D.J., Noguer, M., van der Linden, P.J., Dai, X., Maskell, K., and Johnson, C.A. (eds.), ClimateChange 2001: The Scientific Basis, Cambridge University Press, Cambridge, UK, pp. 892.

Dai, A. G., Meehl, G. A., Washington, W. M., Wigley, T. M. L., and Arblaster, J. M.: 2001a, ‘Ensemblesimulation of twenty-first century climate changes: Business-as-usual versus CO2 stabilization’,Bulletin of the American Meteorological Society 82, 2377–2388.

Dai, A. G., Wigley, T. M. L., Meehl, G. A., and Washington, W. M.: 2001b, ‘Effects of stabilizingatmospheric CO2 on global climate in the next two centuries’, Geophysical Research Letters 28,4511–4514.

de Jager, D., Hendriks, C. M. A., Byers, C., van Brummelen, M., Petersdorff, C., Struker, A. H.M., Blok, K., Oonk, J., Gerbens, S., and Zeeman, G.: 2001, ‘Emission reduction of non-CO2

greeenhouse gases’. Bilthoven, Netherlands, Dutch National Research Programme on Global AirPollution and Climate Change. Report-No.: 953219, available at www.mnp.nl/en

de la Chesnaye, F. C.: 2003, ‘Overview of modelling results of multi-gas scenarios for EMF-21.Presentation at EMF-21 workshop’. Stanford, USA, Energy Modeling Forum, EMF21.

den Elzen, M. G. J.: 2002, ‘Exploring climate regimes for differentiation of future commitments tostabilise greenhouse gas concentrations’, Integrated Assessment 3, 343–359.

den Elzen, M. G. J., and Lucas, P.: 2005, ‘The FAIR model: a tool to analyse environmental and costimplications of climate regimes’, Environmental Modeling & Assessment 10, 115–134.

den Elzen, M. G. J., Lucas, P., and van Vuuren, D.: 2005, ‘Abatement costs of post-Kyoto climateregimes’, Energy Policy 33, 2138–2151.

den Elzen, M. G. J. and Meinshausen, M.: 2005, ‘Meeting the EU 2 ◦C climate target: global and re-gional emission implications’. Bilthoven, Netherlands, RIVM: 44. RIVM report 728001031/2005

Ehhalt, D., Prather, M. J., Dentener, F., Derwent, R. G., Dlugokencky, E., Holland, E., Isaksen, I. S. A.,Katima, J., Kirchhoff, V., Matson, P., Midgley, P., and Wang, M.: 2001, ‘Atmospheric chemistry


and greenhouse gases’, in Houghton, J.T., Ding, Y., Griggs, D.J., Noguer, M., van der Linden,P.J., Dai, X., Maskell, K., and Johnson, C.A. (eds.), Climate Change 2001: The Scientific Basis,Cambridge University Press, Cambridge, UK, pp. 892.

Eickhout, B., den Elzen, M. G. J., and van Vuuren, D.: 2003, ‘Multi-gas emission profiles for stabilisinggreenhouse gas concentrations’. Bilthoven, BA, RIVM. RIVM Report 728001026, available atwww.mnp.nl/en

Enting, I. G., Wigley, T. M. L., and Heimann, M.: 1994, ‘Future emissions and concentra-tions of carbon dioxide: Key ocean/atmosphere/land analyses’, CSIRO Division of Atmo-spheric Research. Research Technical paper no. 31, available at http://www.dar.csiro.au/publications/enting 2001a0.htm

Fearnside, P. M.: 2000, ‘Global warming and tropical land-use change: Greenhouse gas emissionsfrom biomass burning, decomposition and soils in forest conversion, shifting cultivation andsecondary vegetation’, Climatic Change 46, 115–158.

Forest, C. E., Stone, P. H., Sokolov, A., Allen, M. R., and Webster, M. D.: 2002, ‘Quantifyinguncertainties in climate system properties with the use of recent climate observations’, Science295, 113–117.

Fuglestvedt, J. S., Berntsen, T. K., Godal, O., Sausen, R., Shine, K. P., and Skodvin, T.: 2003, ‘Metricsof climate change: Assessing radiative forcing and emission indices’, Climatic Change 58, 267–331.

Graßl, H., Kokott, J., Kulessa, M., Luther, J., Nuscheler, F., Sauerborn, R., Schellnhuber,H.-J., Schubert, R., and Schulze, E.-D.: 2003, ‘Climate Protection Strategies for the 21stCentury. Kyoto and Beyond.’ Berlin, German Advisory Council on Global Change (WBGU),89.

Gregory, J. M., Stouffer, R. J., Raper, S. C. B., Stott, P. A., and Rayner, N. A.: 2002, ‘An observationallybased estimate of the climate sensitivity’, Journal of Climate 15, 3117–3121.

Gritsevskyi, A. and Nakicenovi, N.: 2000, ‘Modeling uncertainty of induced technological change’,Energy Policy 28, 907–921.

Grubb, M. and Ulph, D.: 2002, ‘Energy, the environment, and innovation’, Oxford Review of EconomicPolicy 18, 92–106.

Grubler, A. and Nakicenovic, N.: 2001, ‘Identifying dangers in an uncertain climate’, Nature 412,15–15.

Ha-Duong, M., Grubb, M., and Hourcade, J.: 1997, ‘Influence of socioeconomic inertia and uncertaintyon optimal CO2-emission abatement’, Nature 390, 270–273.

Hansen, J.: 2003, ‘Can we defuse the Global Warming Time Bomb?’ Council on EnvironmentalQuality, Washington: 32.

Hansen, J., Sato, M., Ruedy, R., Lacis, A., and Oinas, V.: 2000, ‘Global warming in the twenty-firstcentury: An alternative scenario’, Proceedings of the National Academy of Sciences of the UnitedStates of America 97, 9875–9880.

Hare, B. and Meinshausen, M.: 2004, ‘How much warming are we committed to and how much canbe avoided?’ PIK Report. Potsdam, Potsdam Institute for Climate Impact Research: 49. No. 93http://www.pik-potsdam.de/publications/pik reports

Harvey, L. D. D.: 2004, ‘Declining temporal effectiveness of carbon sequestration: Implicationsfor compliance with the United National Framework Convention on Climate Change’, ClimaticChange 63, 259–290.

Hill, D. C., Allen, M. R., Gillet, N. P., Tett, S. F. B., Stott, P. A., Jones, G. S., Ingram, W. J., andMitchell, J. F. B.: 2001, ‘Natural and anthropogenic causes of recent climate change’, in India, M.B. and Bonillo, D. L. (eds.), Detecting and Modelling Regional Climate Change, Springer-Verlag,Berlin/Heidelberg, Germany.

Hohne, N., Galleguillos, C., Blok, K., Harnisch, J., and Phylipsen, D.: 2003, ‘Evolution of commit-ments under the UNFCCC: Involving newly industrialized economies and developing countries.’


Berlin, Federal Ministry of the Environment, Nature Conservation and Nuclear Safety: 87. Report-No.: UBA-FB 000412.

Houghton, J., Meira Filho, L., Bruce, J., Lee, H., Callander, B., Haites, E., Harris, N., and Maskell,K., eds: 1994, Climate Change 1994: Radiative Forcing of climate change and the evaluation ofthe 1992 IPCC IS92 emissions scenario, Cambridge University Press, Cambridge.

Houghton, R. A.: 1999, ‘The annual net flux of carbon to the atmosphere from changes in land use1850–1990’, Tellus Series B-Chemical and Physical Meteorology 51, 298–313.

Houghton, R. A. and Hackler, J. L.: 2002, ‘Carbon Flux to the Atmosphere from Land-Use Changes.’A Compendium of Data on Global Change., Carbon Dioxide Information Analysis Center, OakRidge National Laboratory, U.S. Department of Energy, Oak Ridge, Tenn., U.S.A.

IPCC: 1996, Climate Change 1995: the Science of Climate Change. Contribution of WGI to the SecondAssessment Report of the Intergovernmental Panel on Climate Change, Cambridge UniversityPress, Cambridge, UK, p. 572.

IPCC: 2001, Climate Change 2001: The Scientific Basis. Contribution of Working Group I to the ThirdAssessment Report of the Intergovernmental Panel on Climate Change, Cambridge UniversityPress, Cambridge, United Kingdom and New York, NY, USA, 881.

Jiang, K. J., Morita, T., Masui, T., and Matsuoka, Y.: 2000, ‘Global long-term greenhouse gas miti-gation emission scenarios based on AIM’, Environmental Economics and Policy Studies 3, 239–254.

Kirschbaum, M. U. F.: 2003, ‘Can Trees Buy Time? An Assessment of the Role of Vegetation Sinksas Part of the Global Carbon Cycle’, Climatic Change 58, 47–71.

Knutti, R., Stocker, T. F., Joos, F., and Plattner, G. K.: 2003, ‘Probabilistic climate change projectionsusing neural networks’, Climate Dynamics 21, 257–272.

Lashof, D. and Hare, B.: 1999, ‘The role of biotic carbon stocks in stabilizing greenhouse gas con-centrations at safe levels’, Environmental Science & Policy 2, 101–109.

Lean, J., Beer, J., and Bradley, R.S.: 1995, ‘Reconstruction of solar irradience since 1610: Implicationsfor climate change’, Geophysical Research Letters 22, 3195–3198.

Leggett, J., Pepper, W., and Swart, R.: 1992, ‘Emission Scenarios for IPCC: An Update’, in Houghton,J., Callander, B., and Varney, S. (eds.), Climate Change 1992. The Supplementary Report to theIPCC Scientific Assessment, Cambridge University Press, Cambridge, pp. 69–96.

Manne, A. S. and Richels, R. G. U.: 2001, ‘An alternative approach to establishing trade-offs amonggreenhouse gases’, Nature 410, 675–677.

Meinshausen, M.: 2005, ‘What does a 2◦C target mean for greenhouse gas concentration? A briefanalysis based on multi-gas emission pathways and several climate sensitivity uncertainty esti-mates’, in Avoiding Dangerous Climate Change, J.S. Schellnhuber, W. Gamer, N. Nakicenovic,T.M.L. Wigley, and G. Yohe (Eds.), Cambridge Univ. Press

Mitchell, J. F. B., Johns, T. C., Ingram, W. J., and Lowe, J. A.: 2000, ‘The effect of stabilisingatmospheric carbon dioxide concentrations on global and regional climate change’, GeophysicalResearch Letters 27, 2977–2980.

Morita, T., Nakicenovic, N., and Robinson, J.: 2000, ‘Overview of mitigation scenarios for globalclimate stabilization based on new IPCC emission scenarios (SRES)’, Environmental Economicsand Policy Studies 3, 65–88.

Murphy, J. M., Sexton, D. M. H., Barnett, D. N., Jones, G. S., Webb, M. J., Collins, M., and Stainforth,D. A.: 2004, ‘Quantification of modelling uncertainties in a large ensemble of climate changesimulations’, Nature 430, 768–772.

Nakicenovic, N., and Riahi, K.: 2003, ‘Model runs with MESSAGE in the Context of the Further De-velopment of the Kyoto-Protocol’. Berlin, WBGU-German Advisory Council on Global Change:54. Report-No.: WBGU II/2003 available at http://www.wbgu.de/wbgu sn2003 ex03.pdf

Nakicenovic, N. and Swart, R., eds: 2000, IPCC Special Report on Emissions Scenarios, CambridgeUniversity Press, Cambridge, United Kingdom, 612.


Nakicenovic, N., Victor, N., and Morita, T.: 1998, ‘Emissions scenarios database and review ofscenarios’, Mitigation and Adaptation Strategies for Global Change 3, 95–120.

North, G. R., and Wu, Q.: 2001, ‘Detecting Climate Signals Using Space-Time EOF’, Journal ofClimate 14, 1839–1863.

Oppenheimer, M.: 1998, ‘Global warming and the stability of the West Antarctic Ice Sheet’, Nature393, 325–332.

Oppenheimer, M. and Alley, R. B.: 2004, ‘The west antarctic ice sheet and long term climate policy’,Climatic Change 64, 1–10.

Ottinger-Schaefer, D., Godwin, D., and Harnisch, J.: in press, ‘Estimating future emissions andpotential reductions of HFCs, PFCs, and SF6’, Energy Journal.

Prentice, I. C., Farquhar, G., Fasham, M. J. R., Goulden, M. L., Heimann, M., Jaramillo, V. J., Kheshgi,H. S., Le Quere, C., Scholes, R. J., and Wallace, D. W. R.: 2001, ‘The carbon cycle and atmosphericcarbon dioxide’, in Houghton, J. T., Ding, Y., Griggs, D. J., Noguer, M., van der Linden, P. J., Dai,X., Maskell, K., and Johnson, C. A. (eds.), Climate Change 2001: The Scientific Basis, CambridgeUniversity Press, Cambridge, UK, pp. 892.

Pretty, J. N., Ball, A. S., Xiaoyun, L., and Ravindranath, N. H.: 2002, ‘The role of sustainable agricul-ture and renewable-resource management in reducing greenhouse-gas emissions and increasingsinks in China and India’, Philosophical Transactions of the Royal Society of London Seriesa-Mathematical Physical and Engineering Sciences 360, 1741–1761.

Ramaswamy, V., Boucher, O., Haigh, J., Hauglustaine, D., Haywood, J., Myhre, G., Nakajiama, T.,Shi, G. Y., and Solomon, S.: 2001, ‘Radiative forcing of climate change’, in Houghton, J. T.,Ding, Y., Griggs, D. J., Noguer, M., van der Linden, P. J., Dai, X., Maskell, K., and Johnson, C.A. (eds.), Climate Change 2001: The Scientific Basis, Cambridge University Press, Cambridge,UK, pp. 892.

Raper, S. C. B., Gregory, J. M., and Osborn, T. J.: 2001, ‘Use of an upwelling-diffusion energybalance climate model to simulate and diagnose A/OGCM results’, Climate Dynamics 17, 601–613.

Raper, S. C. B., Wigley, T. M. L., and Warrick, R. A.: 1996, ‘Global sea-level rise: Past and future’, inMilliman, J. and Haq, B. U. (eds.), Sea-Level Rise and Coastal Subsidence: Causes, Consequencesand Strategies, Kluwer, Dordrecht, Netherlands, pp. 11–45.

Reilly, J., Prinn, R., Harnisch, J., Fitzmaurice, J., Jacoby, H., Kicklighter, D., Melillo, J., Stone, P.,Sokolov, A. and Wang, C.: 1999, ‘Multi-gas assessment of the Kyoto protocol’, Nature 401,549–555.

Sato, M., Hansen, J., McCornick, M. P., and Pollack, J. B.: 1993, ‘Stratospheric aerosol optiocaldepths’, Geophysical Research Letters 98, 10667–10678.

Schimel, D., Grubb, M., Joos, F., Kaufmann, R., Moss, R., Ogana, W., Richels, R., and Wigley,T. M. L.: 1997, ‘Stabilization of Atmospheric Greenhouse Gases: Physical, Biological and Socio-Economic Implications’. Geneva, IPCC: 56. Technical Paper No.3, available at http://www.ipcc.ch/pub/techpap3.pdf

Smith, J. and Wigley, T.: 2000a, ‘Global warming potentials: 2. Accuracy’, Climatic Change 44,459–469.

Smith, S. and Wigley, T.: 2000b, ‘Global warming potentials: 1. Climatic implications of emissionreductions’, Climatic Change 44, 445–457.

Stott, P. A., Jones, G. S., and Mitchell, J. F. B.: 2003, ‘Do models underestimate the solar contributionto recent climate change?’ Journal of Climate 16, 4079–4093.

Swart, R., Mitchell, J., Morita, T., and Raper, S.: 2002, ‘Stabilisation scenarios for climate impactassessment’, Global Environmental Change 12, 155–165.

Sygna, L., Fuglestvedt, J., and Aaheim, H. A.: 2002, ‘The Adequacy of GWPs as indicators of damagecosts incurred by global warming’, Mitigation and Adaptation Strategies for Global Change 7,45–63.


Thomas, C. D., Cameron, A., Green, R. E., Bakkenes, M., Beaumont, L.J., Collingham, Y. C., Erasmus,B. F. N., de Siqueira, M. F., Grainger, A., Hannah, L., Hughes, L., Huntley, B., van Jaarsveld, A.S., Midgley, G. F., Miles, L., Ortega-Huerta, M. A., Peterson, A. T., Phillips, O. L., and Williams,S. E.: 2004, ‘Extinction risk from climate change’, Nature 427, 145–148.

USEPA: 2003, ‘International Analysis of Methane and Nitrous Oxide Abatement Opportunities.Report to the Energy Modeling Forum, Working Group 21’. Washington, U.S. EnvironmentalProtection Agency http://www.epa.gov/ghginfo/reports/index.htm

van Vuuren, D., den Elzen, M. G. J., Berk, M. M., Lucas, P., Eickhout, B., Eerens, H., and Oostenrijk,R.: 2003, ‘Regional Costs and Benefits of Alternative Post-Kyoto climate regimes’. Bilthoven,RIVM: 117. Report-No: 728001025/2003, available at www.mnp.nl/en

van Vuuren, D. and O‘Neill, B.: submitted, ’The consistency of IPCC‘s SRES scenarios to 1990–2000trends and recent projections’, Climatic Change, 66.

Wassmann, R., Neue, H. U., Ladha, J. K., and Aulakh, M. S.: 2004, ‘Mitigating greenhouse gas emis-sions from rice-wheat cropping systems in Asia’, Environment, Development and Sustainability6, 65–90.

Wigley, T. M. L.: 1991, ‘Could reducing fossil-fuel emissions cause global warming?’ Nature 349,503–506.

Wigley, T. M. L.: 1995, ‘Global mean temperature and sea-level consequences of greenhouse-gasconcentration stabilization’, Geophysical Research Letters 22, 45–48.

Wigley, T. M. L.: 2000, ‘Stabilization of CO2 concentration levels’, in Wigley, T. M. L. and Schimel,D. (eds.), The Carbon Cycle, Cambridge University Press, Cambridge, UK, pp. 258–276.

Wigley, T. M. L.: 2003a, ‘MAGICC/SCENGEN 4.1: Technical Manual’. Boulder, Col-orado, UCAR-Climate and Global Dynamics Division available at http://www.cgd.ucar.edu/cas/wigley/magicc/index.html

Wigley, T. M. L.: 2003b, ‘Modelling climate change under no-policy and policy emissions pathways’.OECD Workshop on the Benefits of Climate Policy: Improving Information for Policy Makers.Paris, France, OECD: 32. OECD: ENV/EPOC/GSP(2003)7/FINAL

Wigley, T. M. L. and Raper, S. C. B.: 1992, ‘Implications for climate and sea level of revised IPCCemissions scenarios’, Nature 357, 293–300.

Wigley, T. M. L. and Raper, S. C. B.: 2001, ‘Interpretation of high projections for global-meanwarming’, Science 293, 451–454.

Wigley, T. M. L. and Raper, S. C. B.: 2002, ‘Reasons for larger warming projections in the IPCCThird Assessment Report’, Journal of Climate 15, 2945–2952.

Wigley, T. M. L., Richels, R., and Edmonds, J. A.: 1996, ‘Economic and environmental choices inthe stabilization of atmospheric CO2 emissions’, Nature 379, 240–243.

Wigley, T. M. L., Smith, S. J., and Prather, M. J.: 2002, ‘Radiative forcing due to reactive gasemissions’, Journal of Climate 15, 2690–2696.

(Received 29 June 2004; in revised form 7 August 2005)

MALTE MEINSHAUSEN , BILL HARE , TOM M.L. WIGLEY , DETLEF … · 2018. 7. 7. · MULTI-GAS EMISSIONS...

Documents

Transcript of MALTE MEINSHAUSEN , BILL HARE , TOM M.L. WIGLEY , DETLEF … · 2018. 7. 7. · MULTI-GAS EMISSIONS...